Privacy Policy

NATEX Work, a service of NATEX AI, Inc.
Effective Date: [[OPEN-01]]
Previous versions: /legal/history
Contact: nshulman@natex.ai · NATEX AI, Inc., a Delaware corporation operated from Michigan. Registered office and address for legal notices: 16192 Coastal Highway, Lewes, DE 19958.

At a glance

This box is a summary. The sections below are the full policy.

• Who's responsible. NATEX AI, Inc. is responsible for your Account and billing information. The owner of a Workspace is responsible for the Content in it, and we handle that Content for the owner. See Section 2 (Our two roles).

• Your data stays in NATEX Work. NATEX Work has its own database and sign-in system, separate from NATEX AI's other products. Your NATEX Work data is used only by NATEX Work. See Section 1 (Who we are and how to reach us).

• What we collect. Your sign-in and Account details, billing details (card data goes to Stripe, never to us), and technical records such as error reports. If you use NATEX Work, we also hold the Content you and your Workspace put in it. See Section 4 (What we collect, where it comes from and why).

• No sale, no ads. We don't Sell or Share Personal Information, and we show no ads. See Section 27 (Notices for readers in particular states).

• AI.

  NATEX does not train models on your content. We ask OpenRouter to route every AI request that uses your plan only to hosts it lists as zero-data-retention, with its zero-data-retention and no-data-collection filters on. If you choose to use your own AI account, those requests go to your provider under your terms with them. A few hosts still run short abuse checks, and web search works differently. Section 6 names them.

  See Section 6 (AI Features and your information).

  AI can be wrong. Check facts and previews before you approve anything.

  Two optional features: Workspace Memory (short summaries of your own work, off until you turn it on) and your own AI account (your content then goes to the provider you chose, under its terms, and not through our routing).

• What other people in your Workspace can see. Every Member except Guests can read the records in a Workspace, including the calendar events and chat messages that a connected app brings in, unless the Workspace restricts them. Mail previews are the exception: only the Member whose mailbox it is can read them, and Admins cannot, unless that Member or an Admin switches on sharing for that mailbox. Sharing is off until someone switches it on; while it is on, every Member except Guests can read that mailbox and send from it, and Guests never can. The Workspace Owner and the Admins can also read Members' AI conversations, and they can see the audit log, roles and last-active times. Only the person who had an AI conversation can delete it. See Section 17 (If you use NATEX Work through someone else's Workspace).

• Export and delete. Settings > Account > Download my data. Settings > Account > Delete my account. Workspace Owners: Settings > Team > Delete this workspace. See Section 23 (Retention Schedule and deletion).

• How long we keep things. It depends on the kind of information. Section 23 (Retention Schedule and deletion) has the table.

• Reach us. nshulman@natex.ai, or the form at /legal/privacy-request.

0. Definitions

A word that starts with a capital letter in this Privacy Policy has the meaning below. The Terms use the same definitions.

Acceptable Use Policy or AUP: Annex A to the Terms.

Acceptance Record: The record we save each time you accept the Terms or the Privacy Policy. It holds the version of each document, the time in UTC, how you accepted (sign-up, create-account, invitation, Google sign-in, magic link or the update dialog), the page, the first 300 characters of your browser's user-agent text, a keyed hash of your IP address, whether you confirmed that you reside in the United States, and, when your network reports one, a two-letter country code. Where your browser reports it, it also holds the version of the arbitration summary that was on your screen and, for the update dialog, the one-line summary of changes that it showed. It never holds your IP address or your date of birth.

Account: Your NATEX Work profile, created after you pass the Age and Country Screen and accept the Terms. Your sign-in identity (your email address and password, Google sign-in or magic link) is held in NATEX Work's own sign-in system, which is separate from the sign-in systems of NATEX AI's other products. Your Account and the Content in your Workspaces are used only by NATEX Work.

Account Data: Personal Information we hold about you as an account holder, apart from Content: your email address, display name and avatar, your sign-in methods, the document versions you accepted, the result of the Age and Country Screen, your security settings (such as two-factor authentication), your role and last-active time in each Workspace, and your plan and billing information.

Add-On: A recurring paid extra you can add to a Paid Plan: AI Power, extra storage (100 GB each) or an extra mailbox. Each Add-On is its own Subscription with its own Billing Period and Renewal Date and is cancelled on its own.

Admin: A Member who has the admin role. The Workspace Owner is also an Admin.

Age and Country Screen: The check at sign-up and, when it is owed, at re-acceptance in which you type your month and year of birth and say where you reside. We use the answer once to decide whether you are an Eligible Person and then discard the month and year.

Agent Action: Any change made or message sent by the Assistant, a Teammate, an Automation, a Plugin, an App, or an API Key (including through the MCP Server), under an Account or a Workspace, whether or not a person clicked at that moment.

Agreement: The Terms together with the documents they incorporate: Schedule 1 (the Plan and Billing Terms), Annex A (the AUP), Annex B (the DPA) where it applies to you, the Seller Payments Terms once we publish them, and the Privacy Policy.

AI Command Pack: A one-time purchase of 600 AI commands that does not renew and starts no Subscription. It is available only on a Paid Plan.

AI Features: Every function that sends text from your Content to a Model Host, or to your Own AI Provider if you use an Own AI Account: the Assistant, search answers, the setup chat, Teammates, summaries of Call Notes, the App and Plugin builder, group planning, web search, and building Workspace Memory and answering questions from it in the "Ask your work" tab of the Ask panel (the "Make changes" tab of that panel is the Assistant). Voice input and live captions are not AI Features; your browser maker's speech service handles them.

AI Vendor: OpenRouter (when it routes a request that uses your plan) and each Model Host.

API: The NATEX Work REST API, version 1.

API Key: A secret key that begins nw_live_, belongs to one Workspace, has a scope, an optional list of allowed collections and an optional expiry date, and lets whoever holds it use the API or the MCP Server for that Workspace.

App: Software generated by the App and Plugin builder or installed from the Gallery that runs inside a Workspace in a sandboxed frame and can use only the permissions that the person who installed it approved.

Approvals: The queue where proposed changes from the Assistant, Teammates, imports and Apps wait for a person to accept or reject them.

Arbitration Agreement: The agreement in Section 31 of the Terms to resolve disputes by individual arbitration, including the class action waiver and the right to opt out.

Assistant: The conversational AI function in NATEX Work that turns your requests into proposals you can review in Approvals.

Authorized User: A Member or Guest who acts in a Workspace for a Customer, including the individual who accepts the Terms for an Organization Customer. An Authorized User who acts for an Organization Customer does not become a party to the Agreement in their own right by doing so.

Auto-Apply: An opt-in setting that an Admin can switch on for one Teammate so that the Teammate's plain create and update changes to ordinary records apply without waiting for a click. It is off by default, switching it on or off is written to the audit log, and each change can be undone.

Automation: A rule you set up in a Workspace that runs by itself when its trigger fires, such as a change to a record or a schedule.

Backup: A rolling copy of the NATEX Work database that we keep apart from the live database for recovery. We keep 14 daily copies and 8 weekly copies. A Backup holds the names and sizes of uploaded files but not the files themselves.

Backup Roll-Off Period: The 60 days after a deletion within which the last copy of the deleted information leaves every Backup.

Billing Owner: The Admin who buys a Paid Plan or an Add-On for a Workspace.

Billing Period: The month or year for which a Subscription is paid. Each Billing Period renews automatically on the Renewal Date until the Subscription is cancelled.

Business Day: A day that is not a Saturday, a Sunday or a US federal holiday.

Call: A real-time audio and video session in NATEX Work between Members, Guests and people who join from a call link. Calls run directly between participants' browsers. If the direct connection fails, a participant can open a backup room, which Google Meet or Zoom hosts if the Workspace has connected that service and which 8x8, Inc. hosts otherwise.

Call Notes: Meeting notes made from the speech recognition built into the browser of each person who adds their speech, shared as one text with everyone in the Call. Call Notes start only when every person in the Call, Guests included, has accepted. A Member can ask an AI Feature to write notes from that text. Live captions are separate: they show a person their own speech to that person only.

Connected Service: A third-party service that you connect to NATEX Work, such as Gmail, Outlook, Google Calendar, Slack, Microsoft Teams, Google Drive, OneDrive, Zoom, Salesforce, an IMAP mailbox, a calendar (ICS) feed or any other app in our catalog.

Consumer Health Data: Personal Information that identifies a person's past, present or future physical or mental health status, as state consumer health data laws define it, that a person enters about themselves, for example workout, meal, habit or journal entries. It does not include Protected Health Information.

Consumer Health Data Policy: The separate page at /legal/consumer-health-data, which is part of the Privacy Policy.

Content: Everything in or created through a Workspace, whoever puts it there: you, your Members and Guests, people who use your Public Pages or Calls, and the Connected Services you connect. It includes records, messages, files, documents, AI conversations, Workspace Memory, Output that you keep, and the copies of these kept in the command log. It does not include Account Data or Usage Data.

Controller: The party that decides why and how Personal Information is processed. It includes a "business" under US state privacy laws. We are the Controller for Account Data, billing, security, abuse, support and product-operations data. The Customer is the Controller for Content.

Counter-Notice: A notice that meets the requirements in Section 23 of the Terms, sent by a person whose Content we removed after a Notice of Claimed Infringement.

Customer: The Individual Customer or Organization Customer that holds a Workspace and, for a Paid Plan, pays for it. A Member of someone else's Workspace is not a Customer for that Workspace and has no billing rights.

Designated Agent: Our copyright agent, registered with the US Copyright Office as DMCA-1075734. Section 23 of the Terms prints the contact block exactly as registered.

Developer Features: The API, the MCP Server, API Keys and Webhooks.

DPA: The data processing addendum in Annex B to the Terms.

Effective Date: The date printed at the top of a document, on which that version takes effect.

Eligible Person: An individual who is at least 18 years old, is at least the age of majority where they reside, and resides in the Territory.

Embargoed Jurisdiction: Any place that is subject to a comprehensive US embargo, as the US Treasury's Office of Foreign Assets Control updates it.

Export: A download of data from NATEX Work: either "Download my data" (your own records as a ZIP of JSON and CSV files, with your uploaded files) or "Download workspace ZIP" (Admins only), also called the Workspace ZIP.

Extra Person: A person added to a Workspace beyond the people included in a Paid Plan, bought as a separate line on the Paid Plan's Subscription. The product calls this a "seat".

Free Plan: The plan that costs nothing, for 1 person, with small usage limits and 0.25 GB of storage. A Workspace moves to the Free Plan when a Trial ends without a purchase or when a Paid Plan ends. The Free Plan is not a Subscription and never renews.

Gallery: The shared catalog where a Member can publish an App, Plugin, theme (called a Look), template or Automation so that other Customers can install it. Only people who are signed in to an Account can see it.

Guest: A person added to a Workspace with the guest role. A Guest joins the conversations that a Member adds them to, and sees other Content only if a Member has shared it with them.

Independent Recipient: A third party that receives information because of how your browser or device works, or because of what you or your Workspace choose, and not because we engage it to process information for us. Examples: your browser maker's speech and push services, Google's STUN servers, MET Norway and the OpenStreetMap Foundation, your mail provider, the vendor of a Connected Service, the owner of a Webhook endpoint, a Third-Party Agent and an Own AI Provider. An Independent Recipient is not a Sub-processor.

Individual Customer: A natural person who creates an Account or buys a Paid Plan for themselves, for any purpose, including freelance work, study or teaching.

Integration Provider: A company we use to connect Connected Services: Composio (Sampark, Inc.) and Nango (Nango Inc).

Intimate Image Request: A request from a person shown in a non-consensual intimate image or video, or from their representative, to remove it. We remove the material within 48 hours of a valid request.

Law Enforcement Guidelines: The policy we follow for requests from governments, in Section 25 of the Privacy Policy.

Legal Hold: A duty to preserve information for a lawsuit or investigation. For our deletion jobs it also means a pause that our Staff set on a named Workspace or Account, on your written request or when the law requires it. The pause covers the deletion jobs that are tied to that Workspace or Account. It does not cover the chat and AI-history period that an Admin chose, general clean-up that is not tied to one Customer, or the Backup Roll-Off Period.

Link Tracking: An option the sender switches on for one email message. Links in that message go through our redirect so that the sender can see the time of each click. The email tells the recipient. It is off by default.

Local Mode: Using NATEX Work without an Account. Your work is stored only in your browser on your device. A page load and an error report still reach our servers.

Local-Only Mode: A setting that stops the Assistant, search and Teammates in NATEX Work from sending your Content to AI. It is different from Local Mode.

Manager: A Member role below Admin with extra rights over assignments and forms.

Material Change: A change to the Terms or the Privacy Policy that reduces your rights, changes the Arbitration Agreement, or lets us use information we already collected in a materially different way. A change of price is not a Material Change; Section S1.6 of Schedule 1 governs it. A Material Change gets 30 days' notice and your re-acceptance.

MCP Server: The Model Context Protocol server that NATEX Work exposes at the API address.

Member: A person with an Account who belongs to a Workspace, whatever their role (Workspace Owner, Admin, Manager or standard member), other than a Guest.

Model Host: A company that runs an AI model for us through OpenRouter. Today these are Microsoft (Azure), Google (Vertex AI), Amazon Web Services (Bedrock), Groq, Deep Infra, Parasail, TypeSafe AI and, for web search only, Perplexity. The Sub-processor List shows the current hosts and is generated from the same source as the Admin AI screen.

Non-User: A person whose Personal Information is in a Workspace but who has not accepted the Terms for that Workspace, such as a contact, a form respondent, a booker, an invoice payer, an email correspondent, a person who joins a Call from a link, a student, a job candidate or the recipient of a tracked link.

Notice at Collection: Section 3 of the Privacy Policy. Sign-up, create-account and first-use screens link directly to it.

Notice of Claimed Infringement: A copyright notice that has the six elements listed in Section 23 of the Terms, sent to the Designated Agent.

On-Device Storage: Information that NATEX Work saves in your browser or on your device: local storage, session storage, IndexedDB and the service-worker cache. Our code does not set cookies. The Stripe and Nango frames that open for payment and app connections may set their own.

OpenRouter: OpenRouter, Inc., through which every AI request from NATEX Work that uses your plan is routed.

Organization Customer: A company, school, nonprofit or other entity that holds a Workspace or buys a Paid Plan. The entity is the Customer and the party to the Agreement. The individual who accepts the Terms for it states that they have authority to bind it, is an Authorized User of it, and does not become a party to the Agreement in their own right by doing so.

Output: Text, code or other content that an AI Feature generates for you.

Own AI Account: An account that you hold with an Own AI Provider and connect to NATEX Work, by saving an API key or, for OpenRouter, by signing in, so that AI Features you run use that account instead of your plan. Only a Member who signs in can connect one.

Own AI Provider: Anthropic, OpenAI, Google or OpenRouter, when you connect an account with it as an Own AI Account. You chose it and you hold the agreement with it. It is an Independent Recipient. It is not a Sub-processor and it is not an AI Vendor.

Paid Plan: Solo, Team, Business or Scale, billed monthly or yearly. The Trial and the Free Plan are not Paid Plans.

Personal Information: Information that identifies, relates to or can reasonably be linked to a person, as the laws that apply define it. Where a breach notification law uses a narrower defined term, we do not use that term; our commitments in the Terms and Privacy Policy apply to a Security Incident whether or not that narrower definition is met.

Plan and Billing Terms: Schedule 1 to the Terms.

Personal Memory: Workspace Memory built for one Member from what that Member can open. Only that Member can read it.

Plugin: An installable, sandboxed extension that adds commands, automation actions, Assistant tools, widgets or settings to a Workspace.

Privacy Center: The in-product module (a module you place on a page, not a Settings page) where a person excludes collections or records from AI, switches on Local-Only Mode, flags Student Data, exports and erases records.

Privacy Policy: The NATEX Work Privacy Policy at /legal/privacy, including the Consumer Health Data Policy.

Processor: A party that processes Personal Information on a Controller's behalf. It includes a "service provider" or "contractor" under US state privacy laws. We are the Processor for Content.

Prohibited Data: Data you must not put into NATEX Work: Protected Health Information of a covered entity or business associate; payment card numbers and security codes; Social Security numbers and other government ID numbers of other people; passwords, private keys and secrets that belong to other people; other people's bank or financial account numbers; biometric identifiers; classified information and technical data controlled under the International Traffic in Arms Regulations or the Export Administration Regulations; and data about children under 13. A Seller's own payment instructions on an invoice or booking page are not Prohibited Data.

Protected Health Information or PHI: Individually identifiable health information held or sent by a HIPAA covered entity or business associate. It does not include an individual's own wellness data.

Public Page: Any page or URL that a person can open without an Account: a form, a booking page, a booking manage link, a hosted invoice, a call link, a Share Link, a public community channel, and a secret calendar feed.

Recording: A video file made from the recorder's own camera and microphone during a Call.

Refund Policy: Section S1.10 of Schedule 1: its six clauses and the rules that follow them.

Renewal Date: The date on which a Subscription renews automatically.

Repeat Infringer: An Account that has received 3 Strikes in 12 months, or 1 Strike for bad-faith or wholesale piracy.

Report: A notice, sent through the in-product "Report" control or by email, of content that a person believes breaks the AUP, the law or their rights.

Restricted Data: Data you may store on your own authority and responsibility, with no assurance from us that NATEX Work fits the rules that apply to it: education records of adult students, employment and candidate data, privileged legal material, financial institution customer information, an individual's own wellness data, and precise location.

Restricted Party: A person on a US sanctions or export denial list, or a person owned or controlled by one.

Retention Schedule: The table in Section 23 of the Privacy Policy that says how long each category of information is kept.

Rights Request: A request to use a right described in Section 26 of the Privacy Policy.

Security Incident: A breach of security that leads to the unauthorized acquisition of, access to, or loss of Content or Account Data held by us, unless, after we look into it promptly and in good faith, we conclude that it is not likely to harm anyone. For the notice to a Workspace Owner or Admin about Content (Section 21.5 of the Terms and Section B7 of Annex B), a Security Incident includes any unauthorized access to a system that holds the Content, whether or not we conclude that it is likely to harm anyone.

Sell and Share: "Sell" means to disclose Personal Information to another party for money or other valuable consideration. "Share" means to disclose Personal Information to another party for cross-context behavioral advertising. We do neither.

Seller: A Customer who uses NATEX Work to invoice, take bookings from or take payment from its own clients.

Seller Payments Terms: The terms that apply to client payments through Stripe Connect. We publish them, and you accept them, only when we turn Stripe Connect on. They are not part of the Agreement at launch.

Sensitive Personal Information: The categories of Personal Information that California law treats as sensitive and that NATEX Work handles: account log-in credentials, the contents of communications we process for you, precise geolocation, and health-related entries.

Service: NATEX Work: the web app, the installable web app, the Public Pages, the Developer Features, the Chrome extension if and when we publish it, our documentation and the marketing site at work.natex.ai.

Share Link: A URL that gives a person without an Account access to a file, folder or document.

Staff: The founder, employees and contractors of NATEX AI, Inc. who are authorized to use our admin tools or the database.

Staff Access Log: The record of Staff reading Content or Account Data through our founder dashboard: reports, copyright records, privacy requests, the customer list and Workspace views. A Workspace's Admins see the entries for their Workspace in the audit log as "NATEX staff". Direct use of the Supabase dashboard, SQL console, a database connection or the service-role key is not recorded by this log, and neither are admin screens that show no Content or Account Data.

Strike: A removal that follows a Notice of Claimed Infringement that substantially complies with Section 23, or a court finding against an Account. A Strike is cleared if a Counter-Notice results in restoration.

Student Data: Records that a teacher or Admin flags as student data, and every assignment submission.

Sub-processor: A company on the Sub-processor List that we engage to process Personal Information in providing the Service. For Content it is our Sub-processor. For Account Data it is our own Processor.

Sub-processor List: The list in Section 8 of the Privacy Policy of each Sub-processor's legal name, purpose, data, location and the date we added it.

Subscription: A Paid Plan or an Add-On, each of which is its own recurring purchase that renews automatically until you cancel it. Extra Persons are lines on a Paid Plan's Subscription.

Team Memory: Workspace Memory that an Admin switches on for a Workspace. It is built only from records that every Member except Guests can already open, and every Member except Guests can read it.

Teammate: A named AI persona (including a scheduled agent) that runs on a schedule or a trigger, can search the web and read Connected Services you allow, and leaves its proposals in Approvals.

Terms: The NATEX Work Terms of Service at /legal/terms, including Schedule 1, Annex A and Annex B.

Territory: The United States, its states, the District of Columbia and its territories (Puerto Rico, Guam, the US Virgin Islands, American Samoa, the Northern Mariana Islands and the US Minor Outlying Islands).

Third-Party Agent: An AI agent or service that holds your API Key or receives data that you route to it.

Third-Party Service: Any product, service, website or content provided by someone other than us, including Connected Services and the services of Independent Recipients.

Trash Window: The 30 days a deleted record stays recoverable before the daily sweep erases it.

Trial: The 14-day period, with no card required, that a new Workspace gets at Team level, for up to 5 people, with set total amounts of usage. Nothing is charged by the Trial. When it ends the Workspace moves to the Free Plan.

Usage Data: Technical and operational information about use of the Service that we generate: usage counters and metering records, error reports, security and rate-limit records, audit logs and the Staff Access Log. It does not include Content.

Visitor: Anyone who opens the marketing site or Local Mode without an Account.

We, us, our: NATEX AI, Inc., a Delaware corporation operated from Michigan.

Webhook: An outbound HTTPS delivery of record events to a URL you choose.

Workspace: The place that holds a Customer's Content, Members, settings and billing.

Workspace Memory: Short plain-text lines that NATEX Work builds from Content, to find the latest on a person or project, each line linked to the records it was built from. It is Personal Memory or Team Memory. Personal Memory is off until the Member turns it on, and Team Memory is off until an Admin turns it on. It is part of Content.

Workspace Owner: The Member who created the Workspace and holds the owner role.

You, your: The person or entity reading the document. For a Customer, it means the Customer. For an Organization Customer, it means the entity and not the individual who accepts or acts for it. For a Member or Guest of someone else's Workspace, it means that person.

Zero-Data-Retention Endpoint: An endpoint that OpenRouter lists as zero-data-retention, meaning one that keeps no prompts or replies beyond serving the request, subject to the exceptions stated in Section 6 of the Privacy Policy.

1. Who we are and how to reach us

1.1 NATEX Work is a service of NATEX AI, Inc., a Delaware corporation operated from Michigan. Registered office and address for legal notices: 16192 Coastal Highway, Lewes, DE 19958. You can reach us about privacy and send a Rights Request at nshulman@natex.ai, or with the form at /legal/privacy-request, which you can use without an Account.

1.2 This Privacy Policy explains what Personal Information we collect through the Service, why we collect it, who receives it, how long we keep it and what choices you have. It does not cover Third-Party Services. A Connected Service, a Third-Party Agent or an Independent Recipient follows its own terms and privacy policy.

1.3 "Including" means including but not limited to.

1.4 NATEX Work has its own database and sign-in system, separate from NATEX AI's other products. NATEX AI uses one Stripe account and one OpenRouter account for all of its products. Your NATEX Work data is used only by NATEX Work.

1.5 The Terms say what you agree to when you use the Service. See the Terms. Section 1 (About these Terms) of the Terms says who the parties are. Section 2 (Our two roles) explains who is responsible for which information.

1.6 Product names are trademarks of their owners. Use here does not imply endorsement or affiliation.

2. Our two roles

2.1 We act in two roles. The role depends on the information.

2.2 As Controller, we decide why and how we use Account Data and Usage Data. Sections 4, 5 and 23 describe that use.

2.3 As Processor, we handle Content for the Customer. The Customer decides what Content goes into the Workspace and why. We process Content to provide the Service and on the Customer's instructions, which are the Terms, the DPA (Annex B to the Terms), the settings the Customer chooses and the way the Customer uses the Service. We do not use Content for any purpose of our own (Section 5). If you are a Member or Guest of someone else's Workspace, the Customer is the Controller of the Content you add there. See Section 19 (Who is responsible for personal information) of the Terms.

2.4 If you are an Individual Customer and you use NATEX Work for personal purposes, you are the Controller of your own Content, and you have your privacy rights directly against us. See Section 26 (Your privacy rights).

2.5 If your information is in someone else's Workspace and you are not a Member or Guest, you are a Non-User. Contact the Workspace's owner. The Workspace named on a Public Page controls the information you submit there. If you tell us which Workspace it is, by email or in the "which workspace or business" field of the form at /legal/privacy-request, we aim to forward your request to the Workspace's Admins within 10 Business Days and help them respond. See Section 16 (If your information is in someone else's Workspace).

3. Notice at Collection

3.1 Who we are. NATEX AI, Inc., a Delaware corporation operated from Michigan. Registered office and address for legal notices: 16192 Coastal Highway, Lewes, DE 19958. Contact: nshulman@natex.ai. The sign-up and create-account screens link to this section.

3.2 What we collect when you create an Account.

3.3 Other points of collection. Each of these has its own section.

• AI Features. The first time you use an AI Feature, text from your Content goes to a Model Host, or to your Own AI Provider if you connected an Own AI Account. See Section 6 (AI Features and your information).
• Connecting a mailbox or an app. The connect screens link to Section 9 (Google user data), Section 10 (Microsoft user data) and Section 11 (Composio, Nango and the Chrome extension).
• Calls and captions. Turning on captions or Call Notes sends your audio to your browser maker's speech service, and Call Notes start only if everyone on the Call accepts. See Section 14 (Calls, Recordings and Call Notes).
• Public forms and bookings. The Workspace named on the page controls the information you submit and gives its own notice. See Section 16 (If your information is in someone else's Workspace).
• Privacy requests. If you send a request through the form or Settings > Account > Privacy requests, we keep your email address, what you asked and our answer, for 24 months after we close the request. We use it only to answer you and to prove that we did.

3.4 Retention. Section 23 (Retention Schedule and deletion) says how long we keep each category.

3.5 No sale. We do not Sell or Share Personal Information. We show no ads.

3.6 Your browser also sends us your IP address and device details with every request. Cloudflare, our hosting company, sees them. We keep a keyed hash of your IP address, not the address, in the Acceptance Record. See Section 22 (Error reports, logs and rate limits).

4. What we collect, where it comes from and why

4.1 The table lists the categories of Personal Information we collect. It uses the category names that California law uses. We collect no biometric information. Content can include information in other categories because the Customer decides what to put in it; the last row covers that.

4.2 Sensitive Personal Information. We use Sensitive Personal Information only to provide the Service you asked for and to keep the Service secure. For that reason we offer no right to limit its use.

4.3 We do not Sell or Share. We do not Sell or Share any category of Personal Information in the table, and we have not done so in the past 12 months. See Section 27 (Notices for readers in particular states).

4.4 Where the recipients are described. Section 7 (Who receives information) groups the recipients. Section 8 (Sub-processor List) names each Sub-processor.

5. How we use information

5.1 We use Personal Information for these purposes:

• (a) to provide the Service: create your Account, run your Workspace, store and show your Content, and run the features you use;
• (b) to authenticate you and keep your Account and the Service secure;
• (c) to bill for Paid Plans, work out tax, send billing notices and handle refunds;
• (d) to detect and stop abuse, spam and illegal use, and to act on Reports and notices;
• (e) to give support that you ask for and to answer privacy requests;
• (f) to keep records that prove your agreement and to defend legal claims;
• (g) to comply with the law and to answer valid legal process; and
• (h) to operate, measure and fix the Service, for example through error reports and usage counters.

5.2 We use your Content only to provide the Service to you and your Workspace, to keep it secure, and as the law requires. We do not use your Content to train or improve AI models, for advertising, or for any purpose of our own, and we do not sell it. We keep limited operating records, such as error reports and usage counters, which the Privacy Policy describes. Section 6 (Your Content) of the Terms says the same.

5.3 We do not show ads. We do not profile you for advertising. We do not Sell or Share Personal Information. We do not train AI models on Content. See Section 6 (AI Features and your information).

5.4 AI Features suggest and people approve. We make no decision with a legal or similarly significant effect about any individual. See Section 6 (AI Features and your information).

5.5 We will not use Personal Information that we already collected in a materially different way without telling you first and, where the law requires it, getting your consent. This includes uses for AI. We will not loosen how we use information we already hold, and we will not apply a new use to it, before that notice. See Section 29 (Changes to this Privacy Policy).

5.6 We email our Staff when someone files a Report or a privacy request, when a purchase comes from outside the United States, and when a system problem needs attention. These emails name the Report or request and the Workspace, not your Content.

6. AI Features and your information

6.1 AI Features send text from your Content to a Model Host. Every request that uses your plan goes through OpenRouter. If you connect an Own AI Account, requests that run on it go to your Own AI Provider instead, as the subsection "Your own AI account" below says. This section says what leaves us, where it goes, what the hosts keep, what we keep, what you can keep out and what an AI Feature cannot promise. The rules for using AI Features are in Section 9 (AI Features) of the Terms.

What each AI Feature sends

6.2 The table lists the AI Features that send text to a Model Host and what leaves us. Prompts can contain Personal Information of Non-Users that is in your Content, such as a contact's name in a record title. We process mail, chat and calendar content only to provide the feature to you. We do not use it for any purpose of our own.

6.3 What you type in the composer is sorted in your browser by fixed rules. Nothing leaves your device until you submit it.

6.4 Teammates and scheduled agents. Scheduled agents and Teammates run without anyone watching. The scheduler looks for due work every 5 minutes, and each Teammate or agent runs on its own schedule or trigger. A scheduled agent pauses after 7 days in which its creator has not been active. We log what a run read as counts of records per collection, the search phrases and web addresses it looked up, and the names of the Connected Service actions it ran. We do not copy the content it read into the log. We also keep the request that summoned a Teammate and what it drafted, and Approvals shows every proposal. Section 10 (Teammates, Automations and actions taken for you) of the Terms says that actions taken for you are your actions. What they propose waits for a person unless Auto-Apply applies:

Scheduled agents and teammates can run while you are away, but what they propose waits in the Approvals queue. The one exception is opt-in and per teammate: a workspace admin can turn on auto-apply, and that teammate’s plain create and update changes to ordinary records are then applied without waiting. Auto-apply never sends, deletes, creates automations or touches invoices, payments, members or other sensitive collections; each change shows in Approvals with an Undo, and turning it on or off is written to the audit log. It is off until an admin turns it on.

Where AI requests go and what the hosts keep

6.5 These statements about AI requests are worded the same in the Terms and in this Privacy Policy.

NATEX Work sends AI requests that use your plan only through OpenRouter, and only to model hosts that OpenRouter lists as zero-data-retention endpoints, meaning endpoints that keep no prompts or replies beyond serving the request, apart from the exceptions below. Every such request asks OpenRouter for both of its filters (zero data retention and no data collection), so if no such host can serve a request it fails instead of going elsewhere. NATEX does not train models on your content. It sends plan requests only to hosts that OpenRouter lists as zero-data-retention endpoints, which keep no prompts or replies to train on, apart from the exceptions below. If you choose to use your own AI account, requests on that account go to your provider instead, and your provider’s terms apply to them.

Those filters do not switch off a host’s own abuse and safety systems. For the hosts we use today:

• Microsoft Azure. Microsoft runs automated abuse monitoring. When it flags a request, a sample of the prompt and the reply may be stored and read by Microsoft reviewers.
• Google Cloud Vertex AI. Google may keep a request in memory for up to 24 hours (caching) and may log prompts to check for abuse, under Google’s own terms.
• Groq. Groq may keep a request for up to 30 days while it troubleshoots an outage or investigates suspected abuse.

OpenRouter keeps request metadata (such as token counts and timing) but not the text of your prompts or replies, and it may sample a small number of prompts to assign anonymous topic categories that are not linked to your account.

We never call the endpoints run by the Chinese model makers themselves (Z.AI, DeepSeek, Moonshot, MiniMax, Alibaba, Baidu and SiliconFlow). Open-weight models from those makers, such as GLM and Qwen, run on US-based hosts.

Web search is different: a short, cleaned search phrase (links, e-mail addresses and long numbers removed, never your workspace) goes to Perplexity through OpenRouter, and Perplexity states that it does not retain or train on API data.

Voice input and live captions use your browser’s own speech recognition, run by your browser maker. Apps you connect (for example Gmail, Slack or Notion) are separate services with their own terms.

6.6 The hosts we use today are Microsoft (Azure), Google (Vertex AI), Amazon Web Services (Bedrock), Groq, Deep Infra, Parasail and TypeSafe AI, plus Perplexity for web search only. Section 8 (Sub-processor List) shows what each one does for us. We reach the Model Hosts and Perplexity only through OpenRouter. We have no separate contract with them. The terms we rely on are OpenRouter's terms and data processing terms and each host's published data statement.

6.7 Most processing happens in the United States. Some Model Hosts and other Sub-processors may process information elsewhere, and we do not promise that it stays in the United States.

6.8 At the hosts named in the exceptions, a person at the host may read a request that the host's systems flag.

6.9 NATEX does not train models on your content. It sends requests that use your plan only to hosts that OpenRouter lists as zero-data-retention endpoints, which keep no prompts or replies to train on, apart from the exceptions stated in Section 6 of the Privacy Policy. If you choose to use your own AI account, requests on that account go to your provider, and your provider’s terms decide what it keeps. We do not keep a license to use your Content to train or improve models, and we do not use data from your Connected Services to train models.

Web search

6.10 A web search sends one short phrase. It has at most 200 characters, and we remove links, email addresses, phone numbers and long number strings from it first. It goes to Perplexity through OpenRouter. Your chat message, your records, your user id, your Workspace id and your email address do not go with it. The phrase can still contain whatever words you or the Assistant chose, such as a name or a topic, so it is not anonymous.

6.11 Perplexity states that it does not retain or train on API data. That is Perplexity's own published statement. We reach Perplexity through OpenRouter and have no separate contract with it. We cannot audit what Perplexity does.

6.12 After a search, our servers fetch the result pages ourselves, using a user agent that names NATEX Work. The sites we fetch see our server's address, not yours. We keep search results in a cache for 24 hours. The Assistant, the "Search the web" mode and Teammates can all run web searches.

AI history, learned preferences and who can read them

6.13 We keep AI conversations until they are deleted, unless an Admin sets 30 days, 90 days or 1 year. The Workspace Owner and the Admins can read Members’ AI conversations in their Workspace. Each person can delete the preferences the Assistant has learned about them. Only the Member who had a conversation can change or delete it. Section 36 (Added terms for Organization Customers) of the Terms lists what Admins can read.

6.14 You can delete one conversation from the assistant's history list ("Delete conversation"), or all of your AI history in a Workspace at Settings > AI > Delete all AI history. Deleting all of it also deletes your Personal Memory in that Workspace and switches it off. Deleting a conversation removes the conversation, its summary, its messages and proposals, and the preferences the Assistant learned from it, at once and with no Trash Window. Records you applied from a proposal stay. Only the person who had a conversation can delete or change it. The Owner and the Admins can read it but cannot delete or change it, and no setting hides it from them. You delete the preferences the Assistant has learned about you one by one at Settings > AI. An Admin sets the retention period at Settings > Admin > Data. A daily sweep at 03:17 UTC deletes older chat messages and AI conversations for good, with no Trash Window; the same period covers Teammate run records, requests and proposals. When you delete your Account, we delete your AI conversations and learned preferences. Copies in the command log expire after 90 days, and copies in Backups are gone within 60 days. See Section 23 (Retention Schedule and deletion).

What you can keep out of AI Features

6.15 An Admin or a Member can turn AI off for a collection. A Member can flag a record "AI never touches this". Records flagged as Student Data are also left out. Before the Assistant, search or a Teammate reads anything, we remove those collections and records on the server. You set these in the Privacy Center module and in Settings > Admin > AI. On Settings > Integrations, each connected app also has an "AI-private" switch. When it is on, NATEX Work does not sync that app's items into the Workspace, and Teammates do not read through that connection. Health collections start excluded in a new Workspace, and any other collection whose name suggests health information is excluded when it is first used. Beyond that, the exclusions apply to what you or an Admin turn on or flag.

6.16 Records flagged as Student Data are excluded from the Assistant, search, Teammates and the Workspace ZIP. The Workspace ZIP leaves out Student Data other than the Admin’s own.

6.17 Local-Only Mode stops the Assistant, search and Teammates in NATEX Work from sending your Content to AI. It does not govern calls you make through the API or Webhooks you configure.

6.18 When Local-Only Mode is on for a Member, the Assistant refuses a request before it reads anything. Local-Only Mode is a setting in Settings > Privacy and, as a second copy of the same setting, in the Privacy Center module. The server enforces it for the Assistant, search, Teammates, the App and Plugin builder, Call Notes summaries, the setup chat and the written reasoning for group plans. Search still finds things by keyword. While any Member of a Workspace has Local-Only Mode on, scheduled agents and Teammate check-ins in that Workspace make no model call.

Accuracy, decisions and corrections

6.19

AI can be wrong. Check facts and previews before you approve anything.

Output can be wrong, incomplete or similar to Output other people get. You are responsible for checking Output before you rely on it or send it.

6.20 AI Features suggest and people approve. Proposals wait in Approvals until a person accepts them, except under Auto-Apply as described above. Changes to your own layout, such as moving a widget, apply at once and can be undone. We make no decision with a legal or similarly significant effect about any individual, and we do not use AI Features to decide eligibility for employment, credit, housing, insurance or education. Customers who use Output to make decisions about people are bound by Section A14 (High-risk decisions about people) of the Terms.

6.21 An AI Feature can state wrong things about real people. If Output about you is wrong, email nshulman@natex.ai with the statement and where you saw it. We treat it as a Rights Request to correct information (see Section 26 (Your privacy rights)). We do not control what a model says. We will look into your report, pass it to the Workspace that holds the Output where needed, and tell you what happened.

Workspace Memory and the Ask panel

6.22 What it is. Workspace Memory is a set of short plain-text lines, for example "Acme's renewal is due on March 3", that help NATEX Work find the latest on a client or project. Each line names the records it was built from. It is built from your mail, chat, calendar, tasks, documents and clients (CRM), and it is part of your Content. The Ask panel, which opens from Ask in the top bar, has two tabs. The "Ask your work" tab answers a question about your work from your records and from this memory. The "Make changes" tab is the Assistant. It reads what the Assistant row in the table in Section 6.2 lists, and it only proposes a change: nothing is written until you confirm it, and you can undo it afterwards.

6.23 It is off until you turn it on. Personal Memory is built for you only after you switch it on at Settings > AI > Workspace memory. Once a day, and when you press "Update now", NATEX Work reads the records that changed since the last update, as the table in Section 6.2 shows, and sends them to an AI Feature that writes the lines. Personal Memory is built only from records you can open. Another Member's mailbox that has not been shared with the Workspace, a direct message or private channel you are not in, and another person's private notes are never read for you. Your own mailboxes, and mailboxes that have been shared with the Workspace for as long as they stay shared, can be read for you.

6.24 What it leaves out. Records in a collection that is turned off for AI, records flagged "AI never touches this", Student Data and health collections that start excluded (Section 6.15) are left out. So are form and assignment submissions, credentials, subscriptions and payments, Members, and our own run records. While Local-Only Mode is on for you, your memory is not built and nothing is sent to a model. "Ask your work" still answers, but only from records on your device, with no AI: it assembles the answer from those records, links each sentence to the record it came from, and says "I don't know" when they do not say. It does the same when you are signed out or when the cited-answer service is not available to you.

6.25 Who can read it. Only you can read your Personal Memory. Other Members and Admins cannot, and no setting changes that. You can edit a line (an edited line is never overwritten by a later update), delete a line, export all your lines as a text file, import lines from such a file, and delete all of them. Switching Personal Memory off stops new updates but keeps the lines until you delete them. "Delete all" at Settings > AI > Workspace memory, and Settings > AI > Delete all AI history, delete your Personal Memory in that Workspace for good, with no Trash Window, and switch it off.

6.26 Team Memory. An Admin can switch on Team Memory at Settings > AI > Workspace memory. It is off by default. It is built only from records that every Member except Guests can already open: shared tasks, projects, deals, contacts, companies, events, documents, announcements, meeting notes, invoices, messages in open channels and shared notes. It uses no direct messages, no private channels and no private notes. It uses mail only from a mailbox that has been shared with the Workspace, and only while that mailbox stays shared: when sharing is switched off or the mailbox is disconnected, every line built from its mail is deleted at once. Every Member except Guests can read Team Memory, and Admins can edit and delete its lines. Switching Team Memory off deletes it at once. Team Memory belongs to the Workspace and goes when the Workspace is deleted. An Admin's Workspace ZIP includes Team Memory, with its lines and the ids of the records each line came from, as a JSON file and a CSV file. A Member's Personal Memory is never in the Workspace ZIP.

6.27 Keeping it in step. At each update, a line is removed when one of its records is deleted, when you can no longer open it, or when it has been turned off for AI. When "Ask your work" quotes a line, it first checks that you can still open every record behind it, and it leaves out any line you could not. So a cited answer comes only from records you can read, whether a model writes it or it is assembled offline from your own records, and it shows only what you may see. If a line you type in one tab reads like it belongs in the other, the panel moves it and tells you. It decides this on your device from the first words, with no model, and sends nothing to do it. A question runs in "Ask your work" at once. A request is placed in the "Make changes" box and is not sent for you, so nothing changes until you confirm a proposal. Every sentence in an answer ends in a numbered link to its record. When nothing in the Workspace supports an answer, it says "I don't know". Memory can be wrong or out of date. Check the records it links to before you rely on it.

6.28 What we keep about asking. We do not save the questions you ask in "Ask your work" or its answers. We keep a count of use and, when a request used your plan, a line in the AI spend ledger (model, tokens, cost and Workspace id, no text), which the Retention Schedule covers. Memory and the "Ask your work" tab share a daily budget for each Workspace, set by the number of seats. When it is used up, memory updates and model-written answers pause until the next day, and search by keyword still works.

6.29 Who processes it. Updates and answers for your Personal Memory go through the same routing as Section 6.5, or to your Own AI Provider if you chose your own account. Team Memory updates always use the plan and never an Own AI Account. The lines themselves are stored in our database, which Supabase hosts.

Your own AI account

6.30 What it is. A Member who signs in can connect an Own AI Account at Settings > AI: an API key from Anthropic, OpenAI, Google or OpenRouter, or a sign-in with OpenRouter. When you choose "my own AI" and the account works, the AI Features that you run (the Assistant, search answers, the setup chat, summaries of Call Notes, the App and Plugin builder, group planning when you are the only one in the request, your Workspace Memory and "Ask your work"), and the Teammates and scheduled agents that you created, send their requests to your Own AI Provider with your key. Web search and the step that routes a request still work as Section 6.2 and Section 6.10 say. Nothing is turned on until you add a key and choose it. You can switch back to your plan at any time.

6.31 What changes for your information. Text from your Content, which can include other people's Personal Information, goes to the company you chose. We are not a party to your agreement with it. Its terms decide what it keeps, whether it reviews requests and whether it trains models. The zero-data-retention routing, the filters and the exceptions in Sections 6.5 to 6.8 do not apply to requests on your Own AI Account, and neither do our statements about hosts. Your Own AI Provider is your recipient, an Independent Recipient. It is not our Sub-processor, we have no contract with it for you, and it does not appear on the Sub-processor List. You are responsible for the terms you accept with it and for your costs, which your provider bills you. Requests on your own account do not use your plan's AI commands.

6.32 Others' information in your requests. Records you can open are not yours alone. In a Workspace that belongs to someone else, think before you send its Content to a provider of your own, and follow the Workspace's rules. The Workspace Owner and the Admins cannot see which Own AI Accounts you connect, but an Admin can switch Own AI Accounts off for the whole Workspace at Settings > AI, and the audit log records the change. While that switch is off, no Member's key is used in that Workspace and our plan models answer under Sections 6.5 to 6.8. A key you saved stays saved, and you can still remove it. Group planning never sends other Members' information to your Own AI Provider: when anyone besides you takes part in a group plan, our plan models write it, and your key is used only when the request holds no one else's context. If the account fails, runs out of credit or is refused, the rest of that request goes to our plan models, under Sections 6.5 to 6.8, and we tell you.

6.33 Your key. We store your key encrypted in our database, bound to you, the Workspace and the provider, and we show only its first characters and last four. No Member, Admin or Staff screen shows it, and the data API cannot read it. We use it only to send your requests and to check that it works. We never print it in a log. If we ever replace the secret that protects saved keys, you will have to paste yours again. Remove a key at Settings > AI and our copy is erased at once. That does not cancel the key at your provider. To stop it working there, delete it in your provider's console. A key made through an OpenRouter sign-in is removed from our records in the same way, and you can end it in your OpenRouter account.

6.34 When it ends. Your saved keys and your Personal Memory in a Workspace are deleted when you leave the Workspace or are removed from it, and your keys and Personal Memory in every Workspace are deleted when you delete your Account. Deleting a Workspace deletes its Team Memory and every key saved in it. Section 23 (Retention Schedule and deletion) says how to export your data and what stays.

7. Who receives information

7.1 In the 12 months before the Effective Date we disclosed Personal Information to the categories of recipients below, for the business purposes shown. We do not Sell or Share Personal Information. See Section 27 (Notices for readers in particular states).

7.2 Independent Recipients are not Sub-processors. An Independent Recipient receives information because of how your browser or device works, or because of what you or your Workspace choose, and not because we engage it to process information for us. Each follows its own terms and privacy policy. The Independent Recipients are:

• Google: connection discovery (STUN) for Calls, Google sign-in, Chrome's speech service, push, Google Calendar when a booker chooses "add to Google Calendar", and Google Meet when it hosts a backup room;
• Microsoft: Edge's speech service, push, and your own Outlook and Teams;
• Apple: Safari's speech service and push;
• Mozilla: the lookup that finds mail server settings for a domain (it receives only the domain), and push;
• MET Norway and the OpenStreetMap Foundation: lookups of an event's location text;
• 8x8, Inc. (meet.jit.si) and Zoom, when they host a backup room (Section 14 says when);
• api.frankfurter.dev (a currency pair) and date.nager.at (a country code), which two Plugins we provide call from our servers; they receive no Content;
• the sites whose web addresses appear in a chat message, which our servers fetch to build a link preview;
• your mail provider and the vendor of any Connected Service;
• Webhook receivers, API Key holders and their Third-Party Agents;
• your Own AI Provider, if you connect an Own AI Account, under your own agreement with it (Section 6 says what we send); and
• any image host that your browser loads a link preview, a mail image or the seller's logo on a hosted invoice from.

7.3 Server-side lookups. The Today weather card is on by default and works only when an event has a location. From our servers, with a user agent that names NATEX Work, we send the event's location text to OpenStreetMap's Nominatim geocoder and the resulting coordinates (to 4 decimals) to MET Norway. Those services see our server's address, not yours. They receive no user id, no Workspace id and no event title. Precise device location is a separate, opt-in setting. It stays in your browser. See Section 19 (Sensitive data).

7.4 Weather data in NATEX Work comes from MET Norway and is licensed under CC BY 4.0 and NLOD 2.0. Place search uses data © OpenStreetMap contributors, available under the ODbL. We provide both as they are. Do not rely on them for decisions about safety. Section 7 (Our materials, brand, feedback and open-source notices) of the Terms repeats this notice.

7.5 A Third-Party Agent that holds your API Key, or that you route data to, is a recipient you chose. It is not our Sub-processor. We are not responsible for what it does with data after we deliver it, and you are responsible for the scopes you give it. Section 12 (API, MCP Server and Webhooks) of the Terms gives the rules for API Keys and Webhooks.

7.6 People and services that you or your Workspace choose also receive information: the people you email, the holders of a Share Link, the owner of a Webhook endpoint, and every Member who can open a Public Page or the Gallery. See Section 18 (Public Pages, the Gallery and community channels).

7.7 Link previews. When a chat message contains a web address, our servers fetch the page to build a preview, using a user agent that names NATEX Work, and keep the preview for 30 days. The site sees our server's address. Each viewer's browser then loads the preview image directly from that site.

7.8 Two Plugins we provide. The currency converter calls api.frankfurter.dev from our servers and sends it a currency pair. The public-holidays widget calls date.nager.at from our servers and sends it a country code. Neither receives Content.

8. Sub-processor List

8.1 This is the Sub-processor List. It shows each Sub-processor's legal name, what it does, the data it handles, where it processes that data, when it is involved, the terms and contract status we rely on and the date we added it. It has two groups: Sub-processors that are always involved, and Sub-processors that are involved only when you use the feature. A company that you connect as your Own AI Provider is not a Sub-processor either, because you chose it and we do not engage it. It is not on this list. The Independent Recipients are not Sub-processors; Section 7 (Who receives information) lists them.

8.2 Contract status. We do not have a separately signed contract with every Sub-processor. For each one we accept its published terms and, where it offers them, its data processing terms. For the Model Hosts and Perplexity, which we reach only through OpenRouter, we accept OpenRouter's terms and data processing terms, and we have no contract of our own with the individual host. We have not signed a data processing agreement with Composio. Composio's standard terms apply. The column "Terms and contract status" says which of these applies to each row.

8.3 We publish the Sub-processor List with each Sub-processor’s legal name, what it does, the data it handles, where it processes that data and the date we added it. We give the Workspace Owner and the Admins an in-app notice in Settings > Privacy at least 30 days before a new Sub-processor starts to process Content. When we do not choose the change ourselves, for example because a Sub-processor changes its own vendors, we give notice as soon as we reasonably can, normally 30 days. If you object within 30 days of the notice, an Admin records the objection in Settings > Privacy with "Object to this Sub-processor". You may then end your plan, and we will refund the unused part of what you prepaid, as Section S1.10 (Refund Policy) of the Plan and Billing Terms describes. We also work with you on any data-protection concern.

8.4 The Sub-processor change notice also appears in Annex B6 (Sub-processors) of the Terms.

9. Google user data

9.1 NATEX Work’s use and transfer to any other app of information received from Google APIs will adhere to Google API Services User Data Policy, including the Limited Use requirements. Section 13 (Connected Services) of the Terms sets out your side.

9.2 What this covers. Google user data means information that we receive from Google APIs. It comes from Google sign-in, and from the Gmail, Google Calendar and Google Drive connections that you set up. Those connections run through our Integration Provider, Composio. The consent screen you see names the app that asks for access.

9.3 What we use it for. We use Google user data to give the features you turn on and for nothing else. Those features are to:

• create and secure your Account when you sign in with Google;
• show your Gmail and Google Calendar items and your Google Drive files in NATEX Work;
• sync them;
• draft messages and send them when you confirm;
• create calendar events, video meetings and replies in your Google account when a feature you use needs one;
• make the changes you make in NATEX Work, such as moving a message to the trash or deleting an event, in your Google account;
• search them; and
• run the Assistant and Search features that you invoke.

We use it only to provide or improve those user-facing features.

9.4 No ads, no sale. We do not use Google user data for advertising. We do not Sell it. We do not give it to data brokers. We do not use it to decide credit or lending.

9.5 No AI training. We do not use Google user data, whether raw, aggregated, anonymized or derived, to create, train or improve a generalized artificial intelligence or machine learning model, ours or anyone else's. We send it only to hosts that OpenRouter lists as zero-data-retention endpoints, which keep no prompts or replies to train on, apart from the exceptions in Section 6. An AI Vendor may process Google user data only to answer the request you made, under the routing and exceptions in Section 6 (AI Features and your information). If you turn on Personal Memory, your synced mail and calendar events are summarized by an AI Feature for you alone. If you connect an Own AI Account, Google user data in a request goes to the Own AI Provider you chose, under that provider's terms and not under Section 6.5. We do not choose it for you. An action that sends or changes something in a Google service needs your confirmation. The Assistant, Teammates and Apps cannot delete anything in a Google service on their own. When you delete an event, or move a message to the trash or archive, in NATEX Work, NATEX Work makes the same change in your Google account.

9.6 People reading it. Our Staff do not read Google user data unless one of these applies:

• you give permission for specific items, for example when you ask for support;
• it is needed for security or to investigate abuse;
• the law requires it; or
• the data is aggregated and used only for our internal operations.

A Model Host's own abuse systems may let its reviewers read a request they flag (Section 6.8). We treat that as the security and abuse exception.

9.7 Transfers. We transfer Google user data to another party only to give or improve the features you use, with your consent; for security or abuse; to comply with law; or in a merger, acquisition or sale of assets, after we get your explicit prior consent. See Section 28 (If our business changes hands).

9.8 What we store, and how to end it. Mail sync stores message details and a preview of up to 280 characters in your Workspace, and calendar sync stores your events. We fetch a full message or attachment only when you open it, through Composio. Booking and invoice emails to your clients go out from your connected Gmail account through Composio. You can revoke our access at any time at myaccount.google.com/permissions. When you disconnect Gmail, Google Calendar or Google Drive at Settings > Integrations, we show you what will be deleted, and then we delete the connection at Composio, the stored sign-in, and what the connection brought in: synced email previews, synced calendar events that you did not edit, and imported files. Things you created or changed yourself, such as drafts, events you made or edited, tasks and notes, stay. An event you edited is unlinked from the connection. A Legal Hold on the Workspace pauses the deletion of the synced items, not the disconnect. You can also ask us to delete Google user data we hold, at nshulman@natex.ai or with the form at /legal/privacy-request; Section 26 (Your privacy rights) says how we answer. When you delete your Account, we delete the mail previews you created in Workspaces that continue, and calendar events and contacts that a mailbox sync created stay as Workspace records. Composio can keep the content of tool calls for up to 1 year after you disconnect (see Section 11 (Composio, Nango and the Chrome extension)). Copies in Backups are gone within 60 days.

10. Microsoft user data

10.1 Microsoft user data means information that we receive from Microsoft APIs through the Outlook, Microsoft 365, OneDrive and Microsoft Teams connections that you set up. Those connections run through Composio. Section 13 (Connected Services) of the Terms sets out your side.

10.2 Purposes. We use Microsoft user data to give the features you turn on. Those features show your mail, calendar, files and chats in NATEX Work, sync them, and search them. They also draft messages and send them when you confirm, and run the Assistant and Search features that you invoke. With your connection we also create calendar events, video meetings and replies in your Microsoft or Zoom account when a feature you use needs one. We make the changes you make in NATEX Work, such as moving a message to the trash or deleting an event, in that account. We ask for the access those features need and use it for nothing else.

10.3 AI use needs your request. We send Microsoft user data to an AI Feature only when you ask the feature to use it, for example when you ask the Assistant about a message. By asking, you give permission for that use. We do not use Microsoft user data to train a generalized AI or machine learning model. We send it only to hosts that OpenRouter lists as zero-data-retention endpoints, which keep no prompts or replies to train on, apart from the exceptions in Section 6 (AI Features and your information). We do not Sell it, and we do not use it for advertising. We do not use it to build a database or a copy of a Microsoft service beyond what the features need. Our Staff read it only in the cases listed for Google user data in Section 9 (Google user data).

10.4 Deleting it. When you disconnect a Microsoft app at Settings > Integrations, we delete the connection, the stored sign-in and what the connection brought in, after showing you the list, as Section 9 (Google user data) describes. When you delete your Account, we delete the mail previews you created in Workspaces that continue and remove your connections. The rules for records already synced and for Backups are the same as in Section 9 (Google user data).

10.5 Revoking access. You can review and remove the access you gave at account.live.com/consent/Manage for a personal Microsoft account and at myapps.microsoft.com for a work or school account.

11. Composio, Nango and the Chrome extension

11.1 Composio and Nango are our Integration Providers. They process the traffic between NATEX Work and the Connected Services you connect, as our Sub-processors, and they hold the sign-in tokens and the keys you paste. Our own database stores an id and a status for each connection, not the token. For apps connected through Composio, read tools run when you ask and tools that write need your confirmation. The Assistant, Teammates and Apps cannot delete anything in a Connected Service on their own. Section 13 (Connected Services) of the Terms sets out your side.

11.2 Composio connects many of the apps you add. It can keep the content of tool calls, which can include message text, for up to 1 year, including after you disconnect an app. That is outside our deletion windows. We have not signed a separate data processing agreement with Composio. Composio's standard terms apply.

11.3 Nango may keep a log of the requests and responses that it proxies, under Nango's own terms. That is also outside our deletion windows. When you disconnect a Nango app, we delete the connection at Nango at once.

11.4 Mailboxes you connect with a password. For a mailbox from another provider, you give us an app password. We store the app password, the username, the server names and the ports encrypted with AES-256-GCM. The email address of the mailbox is stored as an ordinary setting, because it is what you see and what we look the mailbox up by. We use them to read your mail over IMAP and send mail over SMTP. When you disconnect, we remove the password. If we do not know your mail server, we send the domain of your email address, not the address, to Mozilla's mail-settings lookup. Mail sync stores message details and a preview of up to 280 characters.

11.5 Calendar feeds. If you subscribe to a calendar by its web address (an ICS feed), our servers fetch that address every 30 minutes and we keep the address so we can do it. A feed address is often a secret. Keep it as safe as a password.

11.6 We do not use data from your Connected Services to train models. See Section 6 (AI Features and your information).

11.7 NATEX Work includes open-source software. Its notices are at /third-party-notices.txt. We use JSZip under the MIT license and DOMPurify under the Apache License 2.0. Nothing in the Agreement limits the rights you have under an open-source license. Section 7 (Our materials, brand, feedback and open-source notices) of the Terms repeats this notice.

11.8 Browser extension. We have not published a browser extension.

12. Payments and billing

12.1 Stripe processes our payments. When you subscribe, buy an Add-On or buy an AI Command Pack, you pay on a Stripe page. Card details do not touch NATEX Work servers, and we never receive your card number. Section 17 (Invoices, bookings and getting paid by your clients) of the Terms and Section S1.2 (Checkout, consent and what you agree to) of the Terms cover what you agree to when you buy.

12.2 We send Stripe the Workspace id, the plan, the quantity and the id of your confirmation record. You enter your email address, name, billing address (which is required) and payment details on Stripe's page. Stripe sends us the subscription, customer and payment status. We use Stripe Tax to work out any tax from your billing address. Stripe is our Processor for the information we give it. Stripe is also an independent Controller for fraud prevention, compliance and its own legal duties. Stripe's use of information is described in Stripe's privacy policy.

12.3 Stripe keeps the payment records that it must keep, even after you delete your Account or Workspace. When a Workspace is deleted we cancel its Subscription at once, with no refund and no further charge, and delete its Stripe customer record, unless that record has a subscription that belongs to another NATEX AI product. Deleting an Account or a Workspace is not a request for a refund. The delete screens show the period you prepaid but do not offer a refund. If a refund right applies to you, use "End my plan now and get a prorated refund" in Settings > Billing, or ask under the Refund Policy, before you delete. A Workspace that you hand over to another Member keeps its Paid Plan until the paid period ends, and your card stays on the Stripe customer until then. See Plan and Billing Terms.

12.4 Client payments. We do not process client payments at launch. Invoices and booking pages show the payment instructions, such as bank details or a payment link, that the Customer enters. We do not verify them, we are not a party to the payment, and we do not hold funds. Anyone who has an invoice link can see those instructions.

12.5 Billing consent records. Before we send you to Stripe, we save a record of your confirmation. For an automatic renewal you confirm in two separate boxes: one box that holds only your consent to the renewal terms, and one box that accepts the Plan and Billing Terms and the refund policy. The record holds the Workspace, the Member, the kind of purchase, the plan, the interval, the quantity, the price, the version of the Plan and Billing Terms, the exact text we showed you and its SHA-256 code, the label of each of the two boxes, your browser's user-agent text, a keyed hash of your IP address, a two-letter country code, the Stripe session id and the time in UTC. If we cannot save it, we do not charge you. We keep it for 7 years from the purchase, and it stays after you delete your Account or Workspace, because it is evidence of a contract.

12.6 Billing notices. Billing notices are service messages. We send them from noreply@natex.ai through Brevo to every Admin of the Workspace, with a Reply-To address of nshulman@natex.ai. They include the purchase confirmation, renewal reminders, renewal notices, the Trial-ending notice, price-change notices, cancellation confirmations, confirmations that you undid a cancellation, and failed-payment notices. You cannot opt out of them, because they are about your Subscription. Section S1.4 (Renewals and reminders) of the Terms describes the reminders. See Plan and Billing Terms.

12.7 Refund and objection records. When you end a plan with "End my plan now and get a prorated refund", we write a refund record. It holds the Workspace, the Stripe subscription, the reason, the day your plan ended, the Billing Period, the days used and unused, the amount paid and the amount refunded. When an Admin objects to a new Sub-processor, or a person objects to an update of the Terms, we record who objected, when, which notice or version it concerned and any reason given. When we grant a refund right by hand, we record that too. We keep these records for 7 years, and they stay after you delete your Account or Workspace, because they show what we owed and what we paid.

13. Email we send

13.1 Sign-in messages. Confirmation, magic-link and password-reset emails come from noreply@natex.ai. Our sign-in system sends them through Brevo. They hold your email address and a one-time link or code. Nobody reads replies to them; to reach us, write to nshulman@natex.ai. Section 14 (Email, messages and Calls) of the Terms and Section 33 (Notices and electronic communications) of the Terms set the rules for messages.

13.2 Invitations and alerts. We send Workspace invitations and alerts through Brevo. They hold the recipient's email address, and an invitation holds the Workspace's name and the name of the person who invited you. Both carry nshulman@natex.ai as their Reply-To address. We do not use Resend today. If we turn it on for invitations or for relaying mail from a connected mailbox, we first add it to the Sub-processor List and give the notice described in Section 8 (Sub-processor List).

13.3 Messages from your own mailbox. Booking confirmations, reminders and invoice emails to your clients go out from the first Gmail or Outlook account connected to the Workspace, through Composio. The name, email address, booking title and time of your client pass through Composio and the mail provider. If none is connected, the fixed confirmation and reminder texts go instead from NATEX Work's own sender (noreply@natex.ai, through Brevo), shown as " via NATEX Work", with the Workspace Owner's email address as the Reply-To address and a line in the message that says so. We limit this to 100 messages a day for each Workspace and 6 a day for each recipient. Invoice and form emails are not sent this way. Mail you write in the Mail screen can also go out through an IMAP mailbox you connected. These are the Customer's messages, and the Customer is responsible for them.

13.4 Billing notices are service messages. See Section 12 (Payments and billing).

13.5 Marketing email. We send none today. If we start, we will ask for your opt-in first, put an unsubscribe link in every message, include our postal address and keep a list of people who opted out. Billing notices and other service messages are not marketing, and you cannot opt out of them. We send no text messages.

13.6 Link Tracking. A sender can switch on Link Tracking for a message. See Section 16 (If your information is in someone else's Workspace).

14. Calls, Recordings and Call Notes

14.1 How Calls work. A Call runs directly between the browsers of the people in it. Call audio and video do not pass through our servers unless a participant opens a backup room (paragraph 14.2). Each participant's public IP address goes to Google's STUN servers, which help the browsers find each other, and to the other participants. The connection setup messages, which include those addresses, travel through our realtime service. A person who joins from a call link is admitted through a lobby, and we keep the lobby request, which holds the name they gave, for 7 days. The lobby shows a guest the sentence about captions and Call Notes in paragraph 14.5 before they ask to join.

14.2 Backup room. If the direct connection fails, a participant can open a backup room. If the Workspace has Google Calendar connected, we create a Google Calendar event with a Meet link. If it has Zoom connected, we create a Zoom meeting. Otherwise the room is hosted by 8x8, Inc. at meet.jit.si, and the room name contains the Call's secret token. Google, Zoom or 8x8 then receives the audio, video and display names of the people who join, and each decides for itself how to handle them. We offer the backup room on a best-effort basis. Recording and consent duties stay with the host of the Call.

14.3 Recordings. A Recording is made from the recorder's own camera and microphone, on the recorder's own device. Before it starts, the recorder sees a dialog, "Start recording this call?". The dialog says three things. Everyone sees a "Recording" indicator as soon as the recorder starts, and so does anyone who joins later. The Recording uses the recorder's own camera and microphone. The recorder should record only if everyone agreed, because some places require every person's consent. Everyone in the Call, including people who joined from a call link, sees a red bar that reads "Recording" with the recorder's name for as long as the Recording lasts. Someone who joins later sees an alert. Voice rooms cannot record. A Recording needs notice, not each person's acceptance.

14.4 Where a Recording goes. A Recording holds the recorder's own stream. It holds other people's voices only if they come out of the recorder's speakers into the recorder's microphone. When the recorder stops, the recorder chooses to save it to Files, download it or discard it. The choice "Save the recording to Files when I stop" is off by default. Nothing uploads unless the recorder chooses it. A Recording saved to Files carries who recorded it, the start and end times, the duration, the Call's id, the names of the people on the Call when it started, and a flag that says everyone was notified. The recorder is responsible for following the recording consent laws that apply. See Section 14 (Email, messages and Calls) of the Terms.

14.5 Captions and Call Notes. Live captions are each person's own choice. When you turn them on, your browser's own speech recognition, run by your browser maker, receives your audio. Chrome uses Google's speech service, Edge uses Microsoft's and Safari uses Apple's. We do not receive your audio. Call Notes turn speech into text in the same way and then an AI writes notes from the text. The text of what people add travels between the participants through our realtime service and is not stored there. Call Notes start only when everyone on the Call, guests included, accepts. One decline keeps them off for the rest of the Call, and a person who joins later pauses them until that person answers. Everyone sees a "Call Notes on" indicator while they run. A person who turns on Call Notes can also pick up other people's voices from their speakers. We keep each person's answer: who answered, accept or decline, the text we showed, the time, the first 300 characters of the browser's user-agent text and a keyed hash of the IP address, for 3 years. The sentence the lobby shows a guest is: "This call may use live captions or Call Notes. A person who turns them on sends their own microphone audio to their browser maker's speech service. Other people's voices can be picked up from speakers. Call Notes only runs if everyone on the call accepts." A Member can ask an AI Feature to summarize Call Notes, and then up to 30,000 characters of the text go to a Model Host. See Section 6 (AI Features and your information).

Voice input and live captions use your browser’s own speech recognition, run by your browser maker. Apps you connect (for example Gmail, Slack or Notion) are separate services with their own terms.

14.6 Voice input. Push-to-talk voice input in the Assistant's composer uses the same browser speech recognition. We do not run server-side speech-to-text, speaker identification or voice cloning, and we do not make voiceprints. See Section 19 (Sensitive data).

15. On-Device Storage, Local Mode and browser requests

15.1 What we save on your device. Our code does not set cookies. We save On-Device Storage: local storage, session storage, IndexedDB and the service-worker cache. The Stripe and Nango frames that open when you pay or connect an app may set their own cookies. We use no analytics, advertising, session-replay or social-media tools, so there is nothing to opt out of.

15.2 Staying signed in. The session token in local storage keeps you signed in until you sign out. Clearing it. Sign out, then use your browser's "clear site data" for NATEX Work. Signing out alone does not clear IndexedDB, so clear site data if you share the device.

15.3 Local Mode. You can use Local Mode from anywhere, and you do not need an Account. It stores your work only in your browser on your device. A page load still reaches Cloudflare, and an error report still reaches our error-report function (see Section 22 (Error reports, logs and rate limits)). The Local Mode strip reads "Local demo, stored only in this browser · For people 18 and older · Terms · Privacy". To delete your Local Mode work, clear site data for NATEX Work in your browser. Using Local Mode is subject to the Terms, and creating an Account needs you to be an Eligible Person. Section 2 (Accepting these Terms) of the Terms covers Local Mode and the Terms. The weather card does not run in Local Mode.

15.4 Browser requests that show your IP address to others. Some features make your browser contact another company directly, and that company sees your IP address:

• Google's STUN servers, for every Call;
• Google (Meet), Zoom or 8x8, if a backup room is used;
• your browser maker's speech service, for voice input, live captions and Call Notes;
• the push service of your browser maker, if you turn on notifications;
• any https: image host, for a chat link preview, a remote mail image if you allow it, a form logo web address, or the seller's logo on a hosted invoice; and
• Google Calendar, when a booker chooses "add to Google Calendar".

15.5 We serve fonts and whiteboard files from our own servers. Weather and place lookups run from our servers. Visiting NATEX Work does not send your IP address to a font provider.

15.6 Web push. If you turn on notifications, we store your push subscription endpoint until you unsubscribe or your browser's push service rejects it. A notification holds a title and a body of up to 300 characters, such as a message preview. We encrypt it and your browser maker's push service delivers it. That service sees the encrypted notification, your endpoint and the timing. Queued notification text stays in our queue for 7 days. A notification can appear on your lock screen.

16. If your information is in someone else's Workspace

16.1 Customers use NATEX Work to collect and keep information about people who are not Members. If you are one of those people, you are a Non-User. This section is for you.

16.2 The Workspace controls it. The Workspace named on the page controls the information you submit there. Every Public Page except the secret calendar feed, which is a data file, ends with this notice, where {name} is the Workspace's or the Customer's name: Powered by NATEX Work · {name} controls the information you submit · Privacy Where our server does not send a name, the line says "The workspace owner". The Customer is the Controller. We are its Processor. The Customer gives its own notice and answers requests. Each Public Page asks search engines not to list it.

16.3 How to reach the Workspace and us. Contact the Workspace named on the page. You can also email nshulman@natex.ai or use the form at /legal/privacy-request, and tell us which Workspace it is. We aim to forward your request to the Workspace's Admins within 10 Business Days and help them respond. The Workspace Owner can delete the records, and a deleted record stays recoverable for the 30-day Trash Window and is then erased. See Section 26 (Your privacy rights).

16.4 Link Tracking. Link Tracking is an option that the sender switches on for one email message. It is off by default. The box "Track link clicks in this message" is unticked for every message, and the Workspace setting in Settings > Features is also off by default. Links in that message go through our redirect. We store one row for each click: an id, the link's id and the time. We store no IP address and no device information. We do not track whether you opened the email. The email tells you, with this line:

This email uses link tracking: when you click a link, the time of the click is recorded for the sender. Your address, device and location are not.

16.5 If we cannot create the redirect rows, the email goes out unchanged and without that line. Click rows stay until the Workspace or the link is deleted. The redirect limits request rates with a keyed hash of the visitor and the UTC date, and we delete it after 48 hours. We use the address only to limit request rates, we keep only a one-way keyed hash of it, and we delete the hash after 48 hours. Link Tracking is not offered in Local Mode.

16.6 NATEX Work is not directed to children. We do not knowingly collect information from anyone under 13. The Customer is responsible for the information it collects about children, as Section 18 (Your duties to the people whose data you handle) of the Terms explains.

16.7 Mailboxes, calendars and Calls that a Member connects. A Member who connects a mailbox or a calendar represents that the Member has the right to have NATEX Work process those messages and entries for the Member. The Member is responsible for telling the people whose messages, calendar entries and Calls are connected, and for getting any consent the law requires from them, so that we can process those communications to provide the features. See Section 13 (Connected Services) of the Terms.

17. If you use NATEX Work through someone else's Workspace

17.1 If you are a Member or a Guest of a Workspace that belongs to someone else, the Customer controls the Content and decides who can see it. This is who can see what in a team Workspace. The panel "What your workspace admin can see" at Settings > Privacy shows the same list.

Every Member except Guests can see:

• the records, tasks, files and open channels of the Workspace;
• the items that a Connected Service brings into the Workspace as records, other than mail: calendar events, Zoom meetings, Slack and Microsoft Teams messages, issues from GitHub, GitLab and Jira, and files from Google Drive and OneDrive, unless the Workspace restricts them;
• the mail previews and sent mail of a mailbox that has been shared with the Workspace (sharing is off until the Member who connected the mailbox, or an Admin, switches it on, and Guests never see a shared mailbox);
• who is in the Workspace, each person's role and last-active time; and
• which apps are connected, whether they are working, and the account email of each.

Private notes, form submissions and assignment submissions are limited to their author and to Admins and Managers. A Guest sees only the conversations the Guest was added to and the records that are shared with the Guest.

The Workspace Owner and the Admins can also:

• read every Member's AI conversations in the Workspace, but not the preferences the Assistant has learned about a Member, a Member's Personal Memory or a Member's saved AI keys; only the person who had a conversation can delete or change it, and no setting hides it from them;
• read the audit log of changes made in the Workspace;
• download Exports of the Workspace (the Workspace ZIP holds every record that is not deleted, and Team Memory if it is on, but it leaves out private conversations the Admin is not in, the mail previews of other Members' mailboxes that are not shared, and Student Data other than the Admin's own, and its manifest counts what it left out);
• switch a connected mailbox's sharing on or off, including a mailbox that another Member connected; each change is written to the audit log, a shared mailbox shows a "Shared" label to the Members who can use it, and the Member who connected a mailbox is told in NATEX Work, with the name of the person and the time, when someone else switches its sharing on or off, and can switch it off again from that notice; and
• change a Member's role, remove a Member, and manage the apps connected to the Workspace.

Admins cannot read mail previews and sent mail of another Member's mailbox that is not shared, direct messages, group chats and private channels they are not part of, the preferences the Assistant has learned about a Member, a Member's Personal Memory, or the AI accounts a Member saved. Every Member except Guests can read Team Memory if an Admin switches it on.

17.2 Other Members can see your activity through presence and focus-room features. A Guest sees only the conversations the Guest was added to and the records shared with the Guest.

17.3 Entries in the audit log that are labelled "NATEX staff" show when our Staff used our admin tools to look at the Workspace. Our Staff do not read your Content except to give support you ask for, to investigate security or abuse, or to comply with law. Access through our admin tools is recorded in the Staff Access Log, and your Admins can see the entries for your Workspace. Direct access to our database is limited to named, authorized Staff and is not recorded by the application. Section 21 (Security, confidentiality and shared responsibility) of the Terms adds your consent to that access.

17.4 Do not connect a personal mailbox, calendar or app account to a Workspace that belongs to an employer or a school. When a mailbox, calendar, Slack, Microsoft Teams, Zoom, GitHub, GitLab, Jira, Google Drive or OneDrive account is connected, NATEX Work copies what the connection brings in into the Workspace as records. They are email previews of up to 280 characters, calendar events, chat messages, issues and files. Every Member of the Workspace except Guests can open the calendar events, chat messages, issues and files, and the Admins can see that the account is connected. Email previews are different: only the Member who connected the mailbox can open them, and no one else in the Workspace can, Admins included, unless that Member or an Admin switches on sharing for that mailbox. Sharing is off by default. While it is on, every Member except Guests can open that mailbox's mail and send from it, and each send still asks the sender to confirm. Connect a personal account to a Workspace only if you are comfortable with that. Your employer or school is responsible for giving you any notice of monitoring that its law requires. We do not give that notice for it. See Section 36 (Added terms for Organization Customers) of the Terms.

17.5 When you delete your Account, the records you created in a Workspace that continues stay with that Workspace, and your name on them becomes "Deleted user". Chat messages stay with their conversations. Mail previews and private notes you created are erased. See Section 23 (Retention Schedule and deletion).

18. Public Pages, the Gallery and community channels

18.1 Some of what you put in NATEX Work can be seen by people outside your Workspace.

18.2 A link to a Public Page works like a password. Public Pages use long random tokens, and a hosted invoice link has 48 characters. A Share Link set to "Never expires" works until someone revokes it. We delete revoked or expired Share Links 30 days after. Each Public Page asks search engines not to list it.

18.3 Community channels and the Gallery. A community channel that is public shows its posts to anyone with the link. A "Report" control appears on every post of a public community page, where no Account is needed, and on every Gallery item, where you must sign in; see Section 25 (Government requests, safety reports and copyright). The Report control does not appear on Share Links or forms; write to nshulman@natex.ai about those. The Gallery shows the publisher's user id and Workspace id to every signed-in person. An item can contain Personal Information that the publisher chose to include. We may hide, remove, disable or update items and installed copies.

18.4 Apps and Plugins. The permissions screen of an App or Plugin shows the hosts it may call. An installed item that has network permission sends data to the hosts it declares, and only where the Member who installed it approved. The install dialog warns when an item both reads data and can use the network. An App runs in a sandboxed frame that cannot reach the network on its own; its network requests go through our servers to the hosts it declared, and its AI requests go through the Assistant. The hosts an App or Plugin calls are not our Sub-processors. Section 11 (Apps, Plugins and the Gallery) of the Terms and Section 8 (Public Pages, links and what you share) of the Terms set the rules for these items.

19. Sensitive data

19.1 Protected Health Information. NATEX Work is not for Protected Health Information. Do not put the PHI of a covered entity or a business associate into it. We are not a HIPAA business associate, and we sign no business associate agreements. See Section 15 (Prohibited Data and Restricted Data) of the Terms and Section 16 (Professional, regulated and government use) of the Terms, which says what we are not.

19.2 Health entries you make about yourself. Entries such as workouts, meals, habits and journals that you make about yourself are Consumer Health Data. The Consumer Health Data Policy is a separate page that covers them.

19.3 Biometrics. We collect no biometric information. We make no voiceprints and no face templates. Camera and microphone streams are live media between browsers.

19.4 Precise location. Precise device location is off unless you turn on location reminders. It stays in your browser and is compared there with places you tagged. It is not sent to us. The location text of a calendar event is different: it goes from our servers to Nominatim and MET Norway for the weather card. See Section 7 (Who receives information).

19.5 What not to enter. Do not enter Social Security numbers or other government ID numbers of other people. Do not enter other people's bank or financial account numbers, payment card numbers or security codes, or their passwords, private keys or secrets. Do not enter biometric identifiers, classified or export-controlled technical data, or data about children under 13. A Customer's own payment instructions on an invoice or booking page are allowed. These rules are in Section 15 (Prohibited Data and Restricted Data) of the Terms.

19.6 Data you may store on your own responsibility. You may store education records of adult students, employment and candidate data, privileged legal material, financial institution customer information, your own wellness data, and precise location. We give no assurance that NATEX Work fits the rules that apply to them. We do not claim that NATEX Work meets the requirements of HIPAA, FERPA, COPPA or GLBA, and it is not an environment built for them.

19.7 Records flagged as Student Data are excluded from the Assistant, search, Teammates and the Workspace ZIP. The Workspace ZIP leaves out Student Data other than the Admin’s own.

20. Age, children and where you live

20.1 The Age and Country Screen. NATEX Work is for people 18 and older who reside in the United States. At sign-up, and again at a re-acceptance when it is owed, the Age and Country Screen asks you to type your month and year of birth and to say where you reside. Nothing is filled in for you. Your browser sends the month and year to our server over TLS. The server works out a yes or no and then discards the month and year, and it does not log or write them. We store the yes or no, the time, and whether you confirmed that you reside in the United States (yes or no, never an address). A person counts as 18 from the first day of the month after the month of their 18th birthday.

20.2 If you do not pass. You do not get an Account. After an under-18 answer, the form stays closed in that browser for 24 hours. The browser keeps only a timestamp for this, not a date of birth. If a brand-new sign-in identity fails the screen, we delete it if it was created in the last 30 minutes and has no NATEX Work Account. We do not delete an older identity. If the deletion does not happen, the identity stays but has no Account, no Workspace and no data.

20.3 Children and students. NATEX Work is not directed to children. We do not knowingly collect Personal Information from anyone under 18 for an Account, and we do not knowingly collect Personal Information from anyone under 13. If we learn that we hold such information, we delete it. A parent, guardian or school that believes a child has an Account or appears in Content can email nshulman@natex.ai. Teachers and schools may use NATEX Work only for adult Accounts. NATEX Work is not a product for schools that serve children, we offer no school agreement, and we do not claim that it meets FERPA or COPPA.

20.4 Country checks. We use your confirmation of where you reside, the country your network reports when it reports one, and the country of your billing address to check eligibility and to help us follow sanctions law. Stripe collects a billing address for every purchase. If the country your network reports is outside the Territory, we do not create the Account. Checkout refuses an Account that has not confirmed a United States residence, and we flag a completed purchase whose billing country is outside the United States and its territories for review. We email our Staff about it, and we keep the flag (the two-letter country, the Workspace id and the Stripe session id) for 7 years. See Section 27 (Sanctions and export) of the Terms.

20.5 Where you reside. We offer NATEX Work only in the Territory. Section 3 (Who can use NATEX Work) of the Terms sets the eligibility rules. Section 24 (Where information is processed) says more.

21. Security

21.1 We protect Personal Information in these ways:

• We use TLS for information in transit.
• Our infrastructure vendors encrypt data at rest on disk: Supabase disk encryption (AES-256) and Cloudflare R2 (AES-256-GCM, with keys the vendor manages). Disk encryption does not hide Content inside the database from someone who has database access.
• We encrypt stored mailbox (IMAP) app passwords, usernames and server settings, and Webhook secrets, in our application with AES-256-GCM. We also encrypt the AI account keys you save with AES-256-GCM, bound to you, the Workspace and the provider, and keep them where the data API cannot read them.
• We store API Keys only as SHA-256 hashes.
• Row-level security keeps each Workspace's data apart from every other Workspace.
• Actions with real effect need explicit approval.
• Public Pages use long random tokens.
• Two-factor authentication is available, and a Workspace may require it.
• Card data never reaches our servers.

21.2 Backups. We keep a Backup of the NATEX Work database apart from the live database: 14 daily copies and 8 weekly copies, stored in two places, a Supabase Storage bucket and Cloudflare R2. A Backup holds the names and sizes of uploaded files but not the files. Each Backup file is encrypted with AES-256-GCM in our own application before it is stored, and our storage vendors also encrypt their disks. We hold the key only as a server secret, kept apart from the Backups, plus one offline copy that our founder keeps. File names inside a Backup show table names.

21.3 We make no claim of any security certification or audit report. NATEX Work is not a HIPAA, GLBA, FERPA, PCI or FedRAMP environment.

21.4 If something goes wrong. If a Security Incident affects your Content, we will tell the Workspace Owner and the Admins without undue delay and no later than 72 hours after we become aware that it happened, and we will include the information you need for your own notices. If a Security Incident affects Account Data about you, we will tell you without unreasonable delay and no later than 30 calendar days after we become aware of it, and we will notify you and regulators as the law of each affected state requires. Nothing here shortens a deadline that the law gives you or imposes on us.

21.5 Our Staff. Our Staff do not read your Content except to give support you ask for, to investigate security or abuse, or to comply with law. Access through our admin tools is recorded in the Staff Access Log, and your Admins can see the entries for your Workspace. Direct access to our database is limited to named, authorized Staff and is not recorded by the application. Section 21 (Security, confidentiality and shared responsibility) of the Terms, Section B5 (Security measures) of the Terms and Section B7 (Security Incident notice) of the Terms cover the same ground.

21.6 When Staff use the founder dashboard to look at Content or Account Data, the dashboard writes an entry to the Staff Access Log before it returns anything. If it cannot write the entry, it returns nothing. The log covers a Workspace view (Members, connected-account emails, feedback text and recent commands), the customer list, the review and removal of Reports, the copyright ledger and the list of privacy requests. Your Admins see their Workspace's entries in Settings > Admin > Audit log as "NATEX staff". The log does not cover the list of error reports, the list of Legal Holds, or an account deletion that Staff run for a verified request (that is recorded separately, as counts only). A status change on a privacy request, and setting or releasing a Legal Hold, are written to the log after the change. Those entries are internal and are not shown to Admins. Direct use of the Supabase dashboard, the SQL console, a database connection or the service-role key is not recorded by this log. We keep the Staff Access Log for 3 years.

21.7 To report a security problem, email nshulman@natex.ai.

22. Error reports, logs and rate limits

22.1 Error reports. When NATEX Work hits an error it cannot handle, your browser sends a report to our error-report function. It holds the error message, the stack, a source label, the page path without the query string, a platform hint and the Workspace id, if one is loaded. It holds no user id, no IP address and no hash of your IP address. We keep it for 90 days. Local Mode sends error reports too. An error message comes from the code that failed, so in rare cases it can include a few words from the page you were using.

22.2 Feedback. NATEX Work has no feedback form. Feedback you email us stays in our mailbox for as long as we need to answer it. If feedback text reaches our database, we keep it for 730 days.

22.3 Rate-limit counters. To stop abuse of booking, forms, invoices, Share Links, Calls, links, community pages, Reports and privacy requests, we count requests. The part of the counter that identifies a visitor is a keyed hash of the address and the UTC date, never the address. We delete the counters after 48 hours.

22.4 Other operating records. The command log holds every command from NATEX Work, the API or an AI Feature, with its data and the command that undoes it, for 90 days. Usage counters stay for 400 days. The AI spend ledger holds the model, the token count, the cost and the Workspace id, but no text, for 45 days. Records of Webhook deliveries and of messages received from vendors stay for 30 days. Idempotency keys stay for 7 days, API rate counters for 2 days, and the mirror of Stripe events for 90 days. Our uptime checks hold no Personal Information and stay for 35 days. Cloudflare and Supabase keep platform logs under their own terms.

23. Retention Schedule and deletion

23.1 The Retention Schedule below says how long we keep each kind of information. It uses the category names of the table in Section 4 (What we collect, where it comes from and why). "The daily sweep" is a job that runs every day at a fixed time in UTC. The main purge runs at 03:29, the file-removal job at 03:47, and the chat and AI-history job at 03:17. When a row says "Backups", read the Backup Roll-Off Period below.

23.2 Backup Roll-Off Period. A row that you delete leaves the live database at once or after the 30-day Trash Window. A copy can still sit in a Backup. The last copy leaves every Backup within 60 days after the deletion. Uploaded files are not in Backups. Copies in the command log expire after 90 days.

23.3 When you delete something, it leaves the live system at once or after the 30-day Trash Window. Copies in Backups are gone within 60 days. Uploaded files are not in Backups. Copies in the command log expire after 90 days. When you disconnect a Connected Service, we delete what it brought in, after showing you the list. Some information stays where our vendors keep it under their own terms: Composio can keep the content of tool calls for up to 1 year, Brevo keeps delivery logs, Stripe keeps the payment records it must keep, and Supabase and Cloudflare keep platform logs. We keep Acceptance Records, billing consent records and the records of refunds and objections for 7 years as evidence of your agreement and of what we did. The deletion audit row holds counts only, with no name, email or id, for 1,095 days. A Seller’s own Stripe account stays open when a Workspace is deleted.

23.4 Each time you accept the Terms or the Privacy Policy, we save an Acceptance Record. It holds the version of each document, the time in UTC, how you accepted, the page, the first 300 characters of your browser’s user-agent text, a keyed hash of your IP address (a one-way code made with a secret key that only we hold), whether you confirmed that you reside in the United States and, when your network reports one, a two-letter country code. We never save your IP address or your date of birth. We never overwrite an Acceptance Record. We keep Acceptance Records for 7 years from the day you accepted, even if you delete your Account, so that we can prove that you agreed and defend legal claims. When you delete your Account we detach the record from your name and email and keep it under an opaque identifier. Where your browser reports it, the record also holds the version of the arbitration summary that was on your screen next to the box, and the one-line summary of changes shown in an update dialog. A browser that does not report the version leaves it empty. The record never claims that a summary was shown when it was not. Section 2 (Accepting these Terms) of the Terms describes how you accept.

How to export and delete

23.5 These are the places to find each control:

• Download my data: Settings > Account > Download my data (also Settings > Data).
• Delete my account: Settings > Account > Delete my account.
• Delete a Workspace (Workspace Owner only): Settings > Team > Delete this workspace.
• Workspace ZIP (Admins): Settings > Admin > Data > Download workspace ZIP.
• Chat, AI-history and Teammate-run retention (Admins): Settings > Admin > Data.
• Delete one AI conversation: the assistant's history list, "Delete conversation". Delete all your AI history in a Workspace: Settings > AI > Delete all AI history.
• AI learned preferences: Settings > AI. Auto-Apply: the Teammates panel.
• Workspace Memory (switch on or off, update now, edit, export, import, delete all; Admins: Team Memory): Settings > AI > Workspace memory.
• Own AI Accounts for the Workspace (Admins only, a switch that turns them off for everyone): Settings > AI.
• Your own AI account (add, test, choose, remove a key): Settings > AI.
• Privacy requests: Settings > Account > Privacy requests, or the form at /legal/privacy-request.
• Team sharing and what your workspace admin can see (the panel label): Settings > Privacy.
• AI exclusions, Local-Only Mode and the Student Data flag: the Privacy Center module and Settings > Admin > AI.
• Cancel a plan, or end it now with a prorated refund: Settings > Billing.
• Object to a new Sub-processor (Admins): Settings > Privacy.
• Disconnect an app: Settings > Integrations.

23.6 Download my data gives you a ZIP of JSON and CSV files with the records you created and the AI, account and settings information listed in the file, and your uploaded files. It also holds your Personal Memory lines and whether memory is on, and a list of the AI accounts you connected (the provider, the last four characters of the key, the status and dates, never the key itself). It also holds your legal records: every Acceptance Record, the billing consent records of your purchases, your Call Notes answers, the Reports you filed and your privacy requests. It holds up to 50,000 records, 8 MB per file and 40 MB of files, and AI messages and commands run are each capped at 20,000. You can run it 3 times in 10 minutes. The Workspace ZIP is an Admin export of the Workspace's records that are not deleted.

23.7 Deleting your Account. To protect you, we ask for the email address typed in full, a sign-in within the last 10 minutes and, if you use two-factor authentication, a code. You have 5 attempts in 10 minutes. We remove your Stripe and Integration Provider connections first. If either step fails, we delete nothing and you try again. For each Workspace you own that other Members use, you choose to hand it over to a named Member (we suggest the longest-standing other Admin) or to delete it for everyone. A Workspace with no other Member except Guests is deleted with your Account. Your sign-in identity, Account, AI conversations, learned preferences, Personal Memory, saved AI keys, push subscriptions and usage records are deleted. We also delete the Gallery items you published, the calendar feeds you subscribed to, and invitations sent to your address. API Keys you created are revoked. If you cannot sign in, we can delete your Account for you after we confirm control of your email address.

23.8 What stays after you delete your Account. Team Memory stays with a Workspace that continues, because it is built from the Workspace's shared records. Records you created in a Workspace that continues stay with that Workspace, and your name on them becomes "Deleted user". Chat messages stay with their conversations. Mail previews and private notes you created are erased. Calendar events and contacts that a mailbox sync created stay as Workspace records. Files you uploaded stay unless you tick "Also delete the files I uploaded to workspaces that keep going". A Workspace that you hand over keeps its Paid Plan until the paid period ends.

23.9 Deleting a Workspace. The Workspace Owner types the Workspace's exact name and, if two-factor authentication is on, enters a code. We cancel the Subscription at once with no refund, delete the Stripe customer record, delete the connections at the Integration Providers, delete the Workspace's rows (including its Team Memory and every AI account key saved in it) and remove its files. Public links stop working because the rows behind them are gone. A Trial that ends, or a Paid Plan that you cancel, does not delete a Workspace; Section 20 (Export, deletion and retention) of the Terms explains what happens.

23.10 Admin retention settings. An Admin can set chat, AI-history and Teammate-run records to be deleted after 30 days, 90 days or 1 year. The default is to keep them until they are deleted. A Legal Hold does not pause the period that an Admin sets.

24. Where information is processed

24.1 Our database and file storage run in the United States, on Amazon Web Services in us-east-2 (Ohio), through Supabase. Our serverless functions run at a Supabase location near the caller, and we have not pinned a region for them. We have not set a jurisdiction for our Cloudflare R2 storage, so Cloudflare chooses where it is stored. We give you no choice of region.

24.2 Most processing happens in the United States. Some Model Hosts and other Sub-processors may process information elsewhere, and we do not promise that it stays in the United States. If you connect an Own AI Account, your Own AI Provider decides where it processes your requests, and we make no promise about it.

24.3 NATEX Work is offered only to adults who reside in the United States and its territories. It is not directed to the European Economic Area, the United Kingdom, Switzerland or any other country. You confirm that you reside in the Territory. If we find an Account held by a person who resides outside the Territory we may suspend it after notice and a chance to export your data, where the law allows. Section 3 (Who can use NATEX Work) of the Terms repeats this rule.

24.4 We cannot promise that your information is beyond the reach of United States legal process because it is hosted outside the United States. See Section 25 (Government requests, safety reports and copyright).

25. Government requests, safety reports and copyright

Government requests

25.1 This is our Law Enforcement Guidelines. We follow these ten points:

1. Valid process. Send requests only to nshulman@natex.ai, with the identity of the agency, the legal process and the details that identify the Account or Workspace. We may challenge or narrow a request that is overbroad or defective.
2. What each kind of process gets. We disclose Content only with your consent or on a warrant for Content, except for emergency disclosures and reports to the National Center for Missing and Exploited Children, as the law allows. Points 4 and 7 describe those two exceptions. We give basic subscriber information on a subpoena and other records on a court order.
3. Notice. Where the law allows, we send the agency to the Workspace Owner. We tell the Workspace Owner, or for an Individual Customer the person, before we disclose anything, unless a court order forbids notice or there is an emergency. If notice was forbidden, we give it after the order expires.
4. Emergencies. We disclose in an emergency only on a good-faith belief that there is a danger of death or serious physical injury, as 18 U.S.C. § 2702 allows, and we document the basis.
5. Preservation. We honor a request to preserve records for 90 days.
6. Foreign requests. A request from outside the United States needs United States legal process.
7. Child safety. We report apparent child sexual abuse material to the National Center for Missing and Exploited Children, as 18 U.S.C. § 2258A requires.
8. Transparency. We may publish a transparency note.
9. Civil lawsuits. Requests in civil lawsuits for your Content go to you and not to us, as paragraph 25.2 says.
10. Hosting abroad. We cannot promise that data is beyond United States legal process because it is hosted abroad.

25.2 Requests in civil lawsuits for your Content, such as a subpoena or a discovery request, go to you and not to us. We will not give the contents of your Content to a party in a civil case unless you consent or a court order that the law allows requires it. You are responsible for keeping your own data for your legal duties, including turning off retention settings and exporting your data before anything is deleted. If you ask us in writing to pause our deletion jobs for a named Workspace, we will set a Legal Hold, which our Staff set and record. A Legal Hold does not pause the chat and AI-history period that an Admin sets, our general operational clean-up, or Backups. We keep our own records when we are told of a dispute. Section 22 (Government requests, civil subpoenas and Legal Holds) of the Terms says the same.

Reports, safety reports and copyright

25.3 Reports. A "Report" control appears on every post of a public community page, where no Account is needed, and on every Gallery item, where you must sign in. The reasons are spam or scam, harassment or hateful content, illegal content, copyright infringement, intimate image shared without consent, child safety, and something else. You can add details (up to 2,000 characters) and an email address. A copyright Report needs an email address and points to Section 23 (Copyright) of the Terms. The Report control does not appear on Share Links or forms; write to nshulman@natex.ai about those. We limit each visitor to 10 Reports an hour, using a keyed hash, never the IP address.

25.4 What a Report record holds. The kind of Report, its target, the Workspace, the reason, your details, your email address if you gave one, your Account id if you were signed in, and a 400-character excerpt of what was reported, so that a review still works after the content is removed. Staff review Reports in a queue and get an email when one arrives. A Report for an intimate image or child safety is marked high priority. Staff are emailed at once, and the content is hidden from public view until we review it. A review that dismisses the Report puts the content back. "Remove content" deletes a Gallery item or soft-deletes a community post, which stays recoverable for the 30-day Trash Window. "Dismiss" closes the Report. A removal is recorded as Staff access in the Workspace owner's audit log. We keep an open Report until we review it and a reviewed Report for 3 years after the review.

25.5 Intimate images. If you are shown in an intimate image or video that was shared without your consent, or you represent that person, choose "Intimate image shared without consent" in the Report control, or send an Intimate Image Request to nshulman@natex.ai. The request needs your signature, the location of the material, a brief statement that you believe in good faith that the image or video was published without your consent, with any information that helps us see that it was not consensual, and your contact details. We remove the material within 48 hours of a valid request. We also make reasonable efforts to remove identical copies that we know about on the public parts of NATEX Work.

25.6 Child safety. We report apparent child sexual abuse material to the National Center for Missing and Exploited Children as required by law. We keep the material and the related Account Data for 1 year from the report, and we may preserve and disclose related data without notice.

25.7 Copyright. Send a Notice of Claimed Infringement to the Designated Agent, as Section 23 (Copyright) of the Terms describes. If we remove Content and the person who posted it sends a Counter-Notice, we share the Counter-Notice with the claimant. We keep copyright notices, Counter-Notices and our Strike ledger for 3 years after we resolve them.

26. Your privacy rights

26.1 We apply these rights to everyone in the United States as company policy, wherever you reside. This is not a claim that any one state's law applies to you. Section 19 (Who is responsible for personal information) of the Terms explains who answers a request.

26.2 Your rights. You have the right to:

• know and access the Personal Information we hold about you, and the categories, sources, purposes and recipients described in this Privacy Policy;
• delete it;
• correct it;
• receive it in a portable form;
• appeal if we refuse a request;
• be treated the same whether or not you use a right, including on the Free Plan; and
• opt out of Sell or Share of Personal Information and of targeted advertising. None of these happens, so there is nothing to opt out of. If that ever changes, we will treat a Global Privacy Control signal as a valid opt-out.

26.3 How to use them. In NATEX Work: "Download my data", "Delete my account", the Privacy requests panel (Settings > Account > Privacy requests), editing your name and avatar, and deleting the preferences the Assistant learned about you (see Section 23 (Retention Schedule and deletion) for the paths). You can also use the form at /legal/privacy-request without an Account. For everything else, email nshulman@natex.ai.

26.4 Review and change your information. Settings > Account lets you review and change your display name and avatar. "Download my data" gives you the records you created, the AI, account and settings information listed in the file, and your legal records (Acceptance Records, billing consent records, Call Notes answers, Reports you filed and privacy requests). Email us to correct anything you cannot change yourself.

26.5 Verification. We verify a request by your signed-in session or by a reply from your confirmed email address. A request from the public form is stored as unverified, and we confirm control of the address before we release anything. To delete a Workspace that has other Members, we take extra care: only the Workspace Owner can do it. If you cannot sign in, we can delete your Account for you after we confirm control of your email address. An authorized agent may act for you with your written permission and proof of who they are.

26.6 Timing. We acknowledge your request within 10 Business Days and answer within 45 days. We may extend the answer once by another 45 days, and we tell you why.

26.7 Appeal. If we refuse your request, send an appeal through the form (type "Appeal") or email nshulman@natex.ai with the subject "Appeal". We review it again and answer within 45 days, and we explain the result in writing.

26.8 Request log. We keep a log of requests and our answers for 24 months after we close the request.

26.9 Information in someone else's Workspace. If your information is Content in a Workspace that belongs to someone else, we pass your request to that Workspace's Admins, as Section 16 (If your information is in someone else's Workspace) explains. We cannot decide for the Customer what its Workspace keeps.

26.10 When we may refuse. We may refuse or limit a request where the law allows or requires it, for example to keep records we must keep, to protect another person or to defend a legal claim. We tell you the reason.

26.11 We do not loosen how we use information we already collected, including for AI, without prior notice and, where the law requires, your consent. See Section 5 (How we use information).

27. Notices for readers in particular states

27.1 This section collects statements that readers in particular states expect. Every right it mentions is described in Section 26 (Your privacy rights).

27.2 We do not Sell or Share. We do not Sell or Share Personal Information. We have no actual knowledge that we Sell or Share the Personal Information of anyone under 16. Everyone who has an Account is 18 or older, so we do not Sell or Share the information of anyone under 18 either.

27.3 Sensitive Personal Information. We use Sensitive Personal Information only to provide the Service you asked for and for security, so we offer no right to limit its use.

27.4 California terms. California law uses different words for our two roles. For Account Data, billing, security, abuse, support and product-operations data, we are a "business". For Content, we are a "service provider". Section 4 (What we collect, where it comes from and why) lists the categories we collect, their sources and purposes. Section 7 (Who receives information) lists the categories of recipients. We offer no financial incentive or price difference in exchange for Personal Information.

27.5 Do Not Track and Global Privacy Control. We do not respond to browser Do Not Track signals, because we do not track you across other sites. Because we do not Sell or Share, a Global Privacy Control signal changes nothing today. If that ever changes, we will treat the signal as a valid opt-out of Sell and Share. Stripe, which hosts our checkout pages, and Google, for Google sign-in and call connection discovery, may collect information about your device under their own privacy policies.

27.6 Shine the Light. We do not disclose Personal Information to third parties for their direct marketing.

27.7 Colorado, Connecticut and Virginia. You can appeal a refusal as Section 26 (Your privacy rights) describes.

27.8 Washington and Nevada. The Consumer Health Data Policy covers consumer health data.

28. If our business changes hands

28.1 In a merger, acquisition, financing, reorganization, bankruptcy or sale of assets, your information may transfer to the successor, which must honor this Privacy Policy for information collected under it. We will tell you before it is used in a materially different way. Section 34 (General terms) of the Terms covers assignment.

28.2 We transfer Google user data in such a case only with your explicit prior consent. See Section 9 (Google user data).

29. Changes to this Privacy Policy

29.1 The link "Privacy" appears in the footer of our homepage, in the Legal block at Settings > Help, and on the screens where we collect information, such as sign-up, connecting an app and the privacy request screens. This Privacy Policy is plain text on a web page and you can print it. The Effective Date is printed at the top. We review it at least every 12 months.

29.2 If a change is a Material Change, we email the Account address and show a notice in NATEX Work at least 30 days before the change takes effect. The notice names the change, links to the document and to the version history, and shows until the Effective Date. On the Effective Date, the next time you open NATEX Work, a dialog asks you to accept the updated documents, after the Age and Country Screen if one is owed. The dialog shows a short summary of the change, links to the documents, a link to "Download my data" and a box that starts unticked. If the change affects the arbitration agreement, it also shows the arbitration summary and the link to opt out, and your 30 days to opt out start on the day you accept. You choose "Accept and continue" or "Sign out". You do not have to accept. The dialog also has a button, "I don't accept — cancel my plan". It opens a page where nothing is accepted and where you can download your data, cancel your plan, end it now with a prorated refund, or delete your Account (Section 32.3 of the Terms). Declining to accept is not an opt-out. For a change that is not a Material Change, we add it to the version history and the change applies when we post it.

29.3 We do not loosen how we use information we already collected without prior notice and, where the law requires it, your consent. We publish previous versions with their Effective Dates at /legal/history.

29.4 See also Section 32 (Changes to the Terms) of the Terms.

30. Contact and accessibility

30.1 Write to NATEX AI, Inc., a Delaware corporation operated from Michigan. Registered office and address for legal notices: 16192 Coastal Highway, Lewes, DE 19958. Or email nshulman@natex.ai.

30.2 This Privacy Policy is a web page with headings, landmarks and a language attribute, and you can print it. We will give it, or any other legal document, in another format on request. Email nshulman@natex.ai. We aim to make NATEX Work work with assistive technology. We have not certified conformance to WCAG. Report a barrier to nshulman@natex.ai. Section 5 (The Service and how it changes) of the Terms covers accessibility.

30.3 To send a copyright notice, see Section 23 (Copyright) of the Terms. For a Rights Request, see Section 26 (Your privacy rights). To report a security problem, email nshulman@natex.ai. Section 33 (Notices and electronic communications) of the Terms covers legal notices to us.