PamAI Privacy Policy

1. Who We Are

PamAI is operated by Frame AI Transformation Agency Ltd, a company registered in England and Wales, United Kingdom (company number 17518894). Frame AI Transformation Agency Ltd is the controller of the personal data processed by PamAI. We handle personal data under the UK General Data Protection Regulation (UK GDPR) and the Data Protection Act 2018. Our supervisory authority is the UK Information Commissioner's Office (ICO); see Section 14.

• Registered address: 167-169 Great Portland Street, 5th Floor, London W1W 5PF, United Kingdom
• Privacy and data-protection requests: amir@trypamai.com
• General support: amir@trypamai.com

2. What PamAI Does

PamAI is an AI-native project-management product. Its agent helps you manage projects, conversations, files, schedules, connected sources and related work context. PamAI is available as a web app and, in beta, as a desktop app for Mac ("the Desktop app"). The Desktop app works differently from the web app in important ways; Section 6 explains them.

3. Data We Collect

We collect and process:

• Account data: email, name, avatar, timezone, authentication identifiers.
• Workspace and project data: projects, items, milestones, fronts, batches, members and share guests.
• Conversations: chat messages, assistant replies, inbox messages, AI-generated summaries, tool metadata.
• Files and agent memory: files you attach in chat, project files, user memory, project memory, file metadata.
• Voice notes: audio you record for dictation, and the resulting transcript.
• Connected services, if you connect them: Telegram, WhatsApp and Google Calendar (web), and the Desktop app's sources: Notion, ClickUp, GitHub, YouTrack, Granola and local folders on your Mac.
• Desktop device data: a device registration record, app version and sync state for each Mac you use.
• Billing data: plan, subscription status, invoices, purchase history, credit ledger, payment webhook metadata.
• Operational data: request metadata, redacted logs, error reports, security and abuse-prevention metadata.
• AI processing data: prompts, outputs, project context, files and memory needed to answer your requests.

We do not receive or store full card numbers, CVCs or bank details. Payment details are handled by Polar as merchant of record and Stripe as its payment processor.

4. Why We Use Data

We use personal data to:

• Provide the PamAI product and agent.
• Store and retrieve workspaces, projects, messages, files and memory, and sync them to your devices.
• Run the integrations, sources and imports you set up.
• Process subscriptions, credits, invoices, refunds, fraud checks and billing support.
• Secure the service, debug failures, prevent abuse and keep audit records.
• Comply with legal, accounting, tax and regulatory duties.

Our lawful bases under UK GDPR (Article 6) are:

• Contract: to provide the account, workspace, project, agent, integration, billing, support, export and deletion features you request.
• Legitimate interests: to secure PamAI, prevent abuse, debug failures and keep the service reliable, in ways that do not override your rights. This does not include training AI models on your identifiable content, which requires your explicit opt-in (Section 7).
• Consent: where you explicitly opt in, such as allowing AI-provider training on your own interactive conversations, enabling optional integrations, or sending richer Desktop error reports.
• Legal obligation: to keep required tax, accounting, fraud-prevention and compliance records.

5. Sources Of Data

We collect data directly from you, from other PamAI users who invite you to workspaces or share projects with you, from third-party services you choose to connect, from payment and billing providers, and from automatic service logs generated when you use PamAI.

If you join a waitlist or request early access without creating an account, we collect the contact details you submit (such as your email) on the basis of your consent, use them only to contact you about access to PamAI, and delete them when they are no longer needed for that purpose or on your request.

6. The Desktop App (Mac)

What stays on your Mac. The Desktop app keeps your projects' working files, your chats and the full conversation transcripts (including what the agent read and wrote) in the app's folder under ~/Library/Application Support on your Mac. Desktop chats are not uploaded to our servers as conversations. What does reach our servers from a chat is limited to: files the agent creates or changes in a project (which sync, see below), messages you or the agent send to other people through the PamAI inbox, and a short summary of each scheduled routine run.

These local files are not encrypted by PamAI. They are ordinary files and a local database, readable by anyone or any program that can read your user account's files. They are protected only by your Mac's own security, such as your login password and FileVault disk encryption if you have turned it on. If you use Time Machine or another backup tool, these files are copied into your backups, and they stay there under your backup's own retention and encryption settings even after you delete them from PamAI. We recommend turning on FileVault and encrypting your backups.

Passwords and tokens. Your PamAI sign-in and the sign-in tokens for sources you connect are encrypted using macOS's built-in secure storage, whose key is held in your macOS Keychain. They are not stored as plain text.

What syncs to our servers. Project files and memory files sync to our storage (Supabase) so they are available on your other devices, in the web app and to collaborators on shared projects. Content you import from a connected source becomes project files and syncs the same way. Voice notes are sent to our server for transcription (Section 7).

Chat runs on your own AI account. In the Desktop app, the agent runs on your own Claude Code (Anthropic) or Codex (OpenAI) account, which you sign in to on your Mac. Your prompts, the project files the agent reads, and its replies go directly from your Mac to Anthropic or OpenAI under your own agreement with that provider, including its retention and model-training settings. PamAI's zero-data-retention arrangements (Section 7) do not apply to these requests, and we cannot change how those providers handle them. Check your Anthropic or OpenAI account settings if this matters to you. Claude Code and Codex may also keep their own local logs.

Updates and error reports. The Desktop app downloads updates from our download server (hosted on Cloudflare), which sees your IP address and app version. Automatic technical error reports are on by default and can be turned off in the app's settings; see Section 15.

Removing local data. Deleting your PamAI account does not delete files on your Mac. To remove them, quit the app and delete its folder in ~/Library/Application Support, then delete any backups you no longer want.

7. AI Processing

Web app and server-side agent. When you use the web app, and for background jobs such as routines, summaries and notifications, PamAI sends the needed content to AI models through the Vercel AI Gateway. You can pick models from Anthropic (Claude), OpenAI (GPT), Google (Gemini) and some open-weight model families (such as DeepSeek, Kimi and GLM). Whichever model you pick, every request is sent with two settings applied by our server:

• Zero Data Retention (ZDR), on by default: the gateway only uses hosts that have agreed not to retain prompts or outputs after serving the request and not to train on them. ZDR is switched off only for your own interactive conversations if you opt in to training (see below).
• A host allowlist: the gateway may route the request only to the hosts below, and the request fails rather than going anywhere else. Open-weight models are therefore served by these hosts, never by their original developers' own endpoints, and we do not route your data to first-party China AI endpoints.

Training opt-in. You can choose to allow AI providers to train on your own interactive web conversations through a profile setting. It is off by default. Even if you opt in, background and system jobs stay ZDR, and so does any conversation in a project that other people can see, so your choice never applies to other people's content.

Exceptions to the gateway, ZDR or allowlist. Some processing does not follow the rules above:

• Desktop chat runs on your own Claude Code or Codex account, outside our ZDR (Section 6).
• Voice notes (web and Desktop) are sent directly to Groq for transcription. These requests do not go through the gateway and are not covered by a ZDR agreement; Groq's standard API data terms apply.
• Search indexing. To make your chats and project content searchable, we create search embeddings using an OpenAI embedding model through the Vercel AI Gateway with ZDR requested. The host allowlist is not applied to these requests.
• Search ranking. We re-rank search results with the Cohere Rerank model running on Amazon Bedrock, through the Vercel AI Gateway, with ZDR requested.
• Web pages. When the web agent opens a web page, the page is fetched by Cloudflare Browser Rendering. Web searches are run by the AI provider's built-in search tool.

We do not train our own AI models on your content.

8. Web Access And Prompt Injection

The agent can search the web and open web pages, and it reads content from the sources you connect (documents, issues, meeting notes, repositories, email-like messages). Any of this content can contain hidden instructions written by someone else ("prompt injection"). AI models do not reliably ignore such instructions. A malicious page or document could trick the agent into putting information from your project into a web address it then opens, which would send that information to whoever controls that address. We currently do not ask for your approval before each web request. To reduce the risk, connect only sources you trust, be careful about asking the agent to read untrusted pages in projects that hold sensitive information, and tell us if you see the agent behave unexpectedly. We plan to add further safeguards.

9. Sharing And Collaboration

When you invite people to a workspace or share a project (including with guests who join through an email link or share code), they can see the project's content: items, files, memory, and the project's shared context. The agent working in that project, including when a collaborator or guest uses it, can read the same content.

Connected sources in shared projects. If you connect a source (for example Notion, ClickUp, GitHub, YouTrack, Granola or a local folder) to a shared project, text imported from that source becomes part of the project's context, and everyone with access to the project can see it and have the agent use it, even if they have no access to the original service. Only connect sources whose content you are allowed to share with everyone in that project.

10. Connected Services

Google user data (Limited Use). When you connect a Google account (for example, Google Calendar), PamAI accesses only the Google user data needed to provide the features you request, such as your basic profile and email for authentication and your calendar events for scheduling features.

PamAI stores the Google account connection, selected calendar metadata, calendar-derived planning entries and encrypted OAuth tokens only for as long as needed to provide the connected calendar features or meet legal, security or abuse-prevention obligations. Google OAuth tokens are encrypted at rest with AES-256-GCM, Google data is transmitted over HTTPS/TLS, access is limited to authorized systems and personnel, and operational logs are redacted where they may contain personal data.

PamAI's use and transfer of information received from Google APIs to any other app will adhere to the Google API Services User Data Policy, including the Limited Use requirements. Specifically:

• We use Google user data only to provide and improve the user-facing features you requested.
• We do not use Google user data for advertising.
• We do not sell Google user data.
• We do not use Google user data to develop, improve or train generalized or non-personalized AI/ML models.
• Humans do not read your Google user data unless you give consent, it is necessary for security or to comply with law, or the data is aggregated/anonymized and used for internal operations in line with the Limited Use policy.

You can disconnect your Google account in settings. Disconnecting revokes PamAI's access, stops syncing new Google Calendar data, and deletes or deactivates stored tokens; imported calendar entries are deleted where technically possible. Deleting your PamAI account does not currently revoke Google's access token for you. Disconnect Google in settings before deleting your account, or remove PamAI at myaccount.google.com/permissions.

Desktop sources. Connecting Notion, ClickUp, GitHub, YouTrack or Granola uses that service's own sign-in. Some connection steps and change notifications (webhooks) pass through our server; any tokens our server keeps are encrypted with AES-256-GCM. ClickUp connections install webhooks in your ClickUp workspace. Deleting your PamAI account does not currently remove those webhooks or revoke source access; remove PamAI from each service's app or integration settings as well.

11. Sub-processors

We use the providers listed on our public Sub-processors page for hosting, storage, authentication, AI processing, transcription, web fetching, messaging, email, sandboxed code execution, error monitoring and payments.

12. How Your Data Is Stored And Protected

Data is sent over encrypted connections (HTTPS/TLS). Your account data, chats from the web app, project files and memory are stored in Supabase (Postgres database and file storage). That storage is encrypted on disk by our providers, which protects against stolen disks. It is not end-to-end encrypted, and we do not add our own encryption layer to your content. Our servers can read your content in order to run the agent, sync files and provide search, and authorized operators can technically access it. We limit that access to operating, securing and supporting the service. Tokens for connected services are the exception: we encrypt them separately (AES-256-GCM). On your Mac, see Section 6.

13. Who PamAI Is Offered To, And International Transfers

During the beta, PamAI and the Desktop app are offered only to people we invite. We do not restrict access by country, so invited users may be located outside the UK.

We are based in the UK, and several of our providers (listed on the Sub-processors page) process data in the United States and other countries. When personal data leaves the UK, we rely on a transfer mechanism recognised under UK GDPR, such as UK adequacy regulations (including the UK Extension to the EU-US Data Privacy Framework, where the provider is certified), the UK International Data Transfer Agreement, or the UK Addendum to the EU Standard Contractual Clauses, as provided in each provider's data processing terms. You can ask us for details at amir@trypamai.com.

14. Complaints

If you have a concern about how we handle your data, please contact us first at amir@trypamai.com. You also have the right to complain to the UK Information Commissioner's Office (ICO) at ico.org.uk/make-a-complaint or on 0303 123 1113, or to the data protection authority where you live.

15. Logs, Error Reports And Analytics

PamAI does not use third-party product analytics or tracking scripts.

Operational logs and error reports are used to secure and debug the service. Known email, phone, token and short content-preview leak sites are redacted, but redaction is not perfect. Error reports from our server, the web app and the Desktop app go to GlitchTip, an error-monitoring tool we run ourselves on a server rented from Hostinger. They are sent without default personal data such as IP addresses, but a stack trace or error message can occasionally contain a fragment of user content.

The Desktop app sends automatic technical error reports by default; you can turn them off in settings. If you agree, it can also send richer reports for failed commands, which can include the command and its output and therefore project content. Problem reports you submit yourself include the description you write.

AI tracing. To debug and improve the AI agent, our server records traces of each agent turn: timings, errors, which tools ran, and the prompt and output content with known email, phone, token and secret patterns redacted. Redaction is pattern-based and not perfect, so traces can contain parts of your messages and project content. Traces go to a tracing server we run ourselves on a server rented from Hostinger, are not shared with any other company, and are deleted after 1 month. This applies to chats handled by PamAI's servers; chats the Desktop app runs on your own Claude or ChatGPT account are not traced by us.

16. Retention And Deletion

Account, project, chat, file and memory data are kept while needed to provide the service.

Deleting your account. When you delete your account, PamAI deletes your personal workspace and the projects and files stored in workspaces you own. For workspaces owned by someone else, PamAI removes your memberships and your user-scoped files but keeps shared workspace and project content that belongs to other people. If you own a workspace that other people use, deleting your account currently deletes that workspace, including your collaborators' chats, memory and files in it, without warning them; tell them first so they can export what they need.

What deletion does not yet cover (beta). We are being upfront about known gaps we are working on:

• Deleting a project removes it from PamAI, but the files stored for it in our file storage are not yet deleted. They stay until the owning workspace is deleted or until we remove them on request.
• Files synced from the Desktop app are stored by content and are not yet deleted when you delete a project or your account.
• Removing someone from a workspace does not yet delete their private memory files in that workspace.
• Connected services: see Section 10 for Google tokens and ClickUp webhooks.
• Files on your Mac and in your backups are not affected by any server-side deletion (Section 6).

To have all stored files for a deleted project, or all of your Desktop-synced files, removed now, email amir@trypamai.com from your account's email address and tell us which projects or that you want a full deletion. We will confirm when it is done.

Financial and accounting records are retained for 6 years after account deletion, with personal data minimized or anonymized where legally and technically possible.

17. Your Rights

Under UK GDPR you have the right to access your personal data, to have it corrected or erased, to restrict or object to its processing, to receive a copy in a portable format, to withdraw consent at any time where we rely on consent, and not to be subject to decisions based solely on automated processing that significantly affect you. We respond within one month. Depending on where you live, local law may give you similar rights.

You can request data export and account deletion through PamAI or by contacting amir@trypamai.com. Some data may be retained where required for legal, tax, accounting, security or dispute purposes.

18. California Privacy

PamAI does not sell personal data and does not share personal data for cross-context behavioral advertising. PamAI does not use or disclose sensitive personal information for purposes that require a California "limit the use of sensitive personal information" link.

California residents may have rights to know/access, delete, correct, receive a copy of and opt out of certain uses of personal information. PamAI will not discriminate against users for exercising privacy rights.

19. Children

PamAI is for users aged 16 or older. We do not knowingly collect personal data from anyone under 16.

20. Cookies And Local Storage

PamAI uses essential cookies only, such as authentication and session cookies, and stores some data in your browser's local storage (such as unsent message drafts and cached project data). PamAI does not use non-essential analytics cookies, so no cookie-consent banner is used. See our Cookie Notice.

If PamAI adds product analytics, session replay, advertising pixels or tag managers later, we will add cookie consent and update this policy first.

21. Beta

The Desktop app is in beta. Features, providers and data flows may change faster than usual during the beta, and some protections described above are still being built. We will update this policy before a change that materially affects how your data is handled.

22. Changes

We may update this policy when the product, providers, legal requirements or data practices change. Material changes will be reflected with an updated date and appropriate notice.