Privacy Policy
Last updated 17 October 2026. Effective 17 October 2026.
1. Introduction
This Privacy Policy explains how SafePod ("we", "us") collects, uses, discloses, retains and protects personal data when you use the SafePod platform at safepod.ai and its associated applications (the "Service"). SafePod is a platform for running AI agents, each in its own isolated cloud environment.
This policy applies to everyone who creates a SafePod account or visits our website. It does not apply to third-party services you choose to connect to SafePod, which are governed by their own privacy policies. Your use of the Service is also subject to our Terms of Service.
Data controller. SafePod is the controller of the personal data described in this policy. For any privacy question or request, including a request to access or delete your data, contact [email protected].
2. Information we collect
2.1 Information you provide to us
- Waitlist signups. Before you have an account you can ask for a place on the waitlist. We collect your email address and the optional note you leave, and use them only to send one confirmation and to tell you when a place opens. Every confirmation carries a one-click removal link, and we keep the entry until you use it.
- Account information. When you sign in with Google, we receive your email address, name and Google account identifier. If you sign in with a one-time code, we receive your email address.
- Credentials you add. Model provider API keys and the authorizations issued when you connect a third-party service. These are encrypted before storage and are described in section 9.
- Configuration. The agents you create, their settings, and the permissions you grant each of them.
- Correspondence. Messages you send us for support or other enquiries.
2.2 Information created when you use the Service
- Workspace content. Files your agents create, read or modify. This content is the product; we store it so your work is available when you return.
- Conversations. The prompts you send, the agents' responses, and the tool activity associated with them.
- Activity records. A timeline of what each agent did — the action, its target and its outcome, including actions that were refused. For actions sent to a connected third-party service, we record a hash of the action input rather than the input itself, so the body of a message or document does not enter the activity record.
2.3 Information from services you connect
If you connect a third-party service, we receive data from that service on your instruction and on your behalf. Google user data is covered separately in section 5. Other connectors (for example Slack, Telegram, Lark, Feishu or Microsoft OneDrive) return only the data required by the specific action you granted.
2.4 Information we collect automatically
- Server logs. Timestamps, request paths, HTTP status codes, error reasons, and the IP address a request came from, kept for reliability, security and abuse prevention.
- Device and browser information sent by your browser as part of ordinary requests.
- Anti-bot signals, when you use the waitlist form. The form is protected by Cloudflare Turnstile, which receives your IP address, TLS fingerprint, User-Agent header, and the site key and origin of the page. Turnstile runs in invisible mode, so there is nothing for you to click — which is precisely why it is named here.
We do not collect precise geolocation, biometric data, or special categories of personal data, and we do not ask you to provide them.
3. How we use your information
| Purpose | Data used |
|---|---|
| Provide and operate the Service — run your agents, store your workspaces, return your work | Account information, configuration, workspace content, conversations |
| Authenticate you and keep your session secure | Account information, session cookies, server logs |
| Carry out the actions you grant an agent against a service you connected | Credentials you add, data from connected services |
| Give you an auditable record of what your agents did | Activity records |
| Maintain security, investigate abuse, and debug faults | Server logs, activity records |
| Send operational and service email, such as sign-in codes and account notices | Account information |
| Hold a place on the pre-launch waitlist and tell you when one opens | Waitlist signups |
| Tell a person from a bot on the waitlist form | Anti-bot signals |
| Comply with legal obligations and enforce our terms | As required in the circumstances |
We do not:
- sell your personal data, or share it for cross-context behavioural advertising;
- use your data for advertising, or serve advertisements;
- use your workspace content, your conversations, or data received from connected services to train any machine-learning model, ours or anyone else's;
- access your workspace content for any purpose other than operating the Service for you, and we do not grant staff routine access to it.
We do not use your personal data for automated decision-making that produces legal or similarly significant effects concerning you.
4. Legal bases for processing
If you are in the European Economic Area, the United Kingdom or Switzerland, we rely on the following legal bases under the GDPR:
- Performance of a contract — to provide the Service you have signed up for, including running your agents and storing your workspaces.
- Consent — to connect a third-party account and access data from it. You may withdraw this consent at any time by disconnecting the service, as described in section 10.
- Legitimate interests — to keep the Service secure, prevent abuse, and diagnose faults, balanced against your rights and freedoms. This is also the basis for the anti-bot check on the waitlist form.
- Legal obligation — where we are required to retain or disclose data by law.
5. Google user data
If you connect Gmail or Google Drive, SafePod requests the permissions listed below at the moment you connect that specific service. If you never connect it, we never request them. Signing in to SafePod with Google alone does not grant any of them.
| Google permission | Why SafePod requests it | Where the data goes |
|---|---|---|
https://www.googleapis.com/auth/gmail.readonly |
Search mail threads, read a thread you ask an agent to read, and read an attachment on a thread the agent has been granted. Only an agent you granted the matching action can do this. | Into that agent's conversation and, if the agent writes it out, that agent's workspace. Not into the activity record, which stores an input hash only. |
https://www.googleapis.com/auth/gmail.compose |
Create a draft in your Gmail Drafts folder. Nothing is sent: SafePod contains no send-mail code path, no send action, and no call to a Gmail send endpoint. You review and send drafts yourself in Gmail. Google documents this permission as also permitting sending, and no narrower Google permission can create a draft, so the authorization you grant is broader than what SafePod is able to do with it. | The draft is created in your mailbox. SafePod stores the draft identifier it received, not the draft body. |
https://www.googleapis.com/auth/drive.readonly and
https://www.googleapis.com/auth/drive |
Search your Drive and read the files you ask an agent to work with. | Into that agent's conversation and workspace, as above. |
https://www.googleapis.com/auth/drive.file |
Upload a file an agent produced into your Drive, and create a folder to hold it. | The file is written to your Drive. SafePod keeps the resulting file identifier. |
Limited Use
SafePod's use and transfer to any other app of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements.
In particular, we do not transfer Google user data to third parties except as necessary to provide or improve the features you are using, to comply with applicable law, or as part of a merger or acquisition with notice to you; we do not use it for advertising; we do not allow humans to read it except with your affirmative consent for specific messages, where necessary for security purposes such as investigating abuse, to comply with applicable law, or where the data has been aggregated and de-identified; and we do not use it to develop, improve or train generalised or non-personalised AI or machine-learning models.
6. How we share information
We do not sell personal data. We share it only in the following circumstances.
6.1 Service providers
- Google Cloud Platform — compute, managed database and object storage.
- Cloudflare — DNS and edge protection for SafePod's own domains.
- An email delivery provider — to send sign-in codes and operational notices over authenticated, encrypted SMTP.
- Cloudflare Turnstile — to keep automated signups off the waitlist form. Cloudflare acts as our processor for this, and states that the purpose of collecting the signals "is not to identify, profile or target any individuals but solely to detect and block bots", and that it cannot directly identify an individual from them. Cloudflare separately acts as its own controller when it uses the signals to improve Turnstile. See the Turnstile Privacy Policy. That document does not state a retention period; for questions Cloudflare directs people to [email protected].
These providers process data on our instructions, under written terms, and only to provide their service to us.
6.2 Services you direct us to use
- The model provider you choose. Your prompts and the context an agent sends are transmitted to the provider whose API key you added — for example Anthropic, OpenAI, DeepSeek, GLM, Kimi, Qianwen, MiniMax, OpenRouter, or an OpenAI-compatible endpoint you configure. SafePod does not choose this for you, and that provider's own terms govern what it does with the traffic.
- The connectors you enable. Requests are made to that provider's API — for example Google, Slack, Telegram, Lark, Feishu or Microsoft — on your behalf and within the permissions you granted.
6.3 Legal and corporate
- Legal requirements. We may disclose data where we believe in good faith that we are required to do so by law, or to protect the rights, property or safety of SafePod, our users or the public.
- Business transfers. If SafePod is involved in a merger, acquisition or sale of assets, personal data may be transferred as part of that transaction. We will notify you before your data becomes subject to a different privacy policy.
7. International data transfers
SafePod operates on Google Cloud in the United States (region us-central1). If you access the
Service from outside the United States, your personal data will be transferred to and processed in the
United States, which may not provide the same level of data protection as your jurisdiction. Where required,
we rely on the European Commission's Standard Contractual Clauses, or an equivalent transfer mechanism, as
the legal basis for such transfers. Dispute resolution is addressed in the
Terms of Service.
8. Cookies and similar technologies
SafePod uses strictly necessary cookies only. These are HttpOnly, Secure session
cookies that keep you signed in and protect against cross-site request forgery; one additional short-lived
cookie is set if you open a shared workspace link. We do not use advertising cookies, cross-site tracking
technologies, or third-party marketing pixels, and we therefore do not present a cookie consent banner for
non-essential cookies, because we set none.
The waitlist form loads Cloudflare Turnstile, which may set its own strictly necessary cookies to distinguish people from bots. These are not advertising or tracking cookies, and Cloudflare states the signals are used solely to detect and block bots. The Turnstile Privacy Policy does not enumerate the cookies it sets; we link it rather than paraphrase it.
Our application includes optional product analytics that is disabled unless explicitly configured for a deployment. Where it is enabled, it records usage events such as which screens were opened; it does not record your workspace content, your conversations, or data received from connected services.
9. Security
We maintain technical and organisational measures appropriate to the risk, including the following.
- Credentials are kept out of the agent. A running agent receives the placeholder string
safepod-proxy-managedinstead of your key or token. The real credential is attached to the outbound request outside the agent's sandbox, by a component the agent cannot reach. An agent tricked into revealing its own credentials reveals a placeholder. - Each agent is isolated. Every running agent executes in a gVisor-sandboxed container with its own filesystem, its own process space, and a network policy that denies traffic by default.
- Encryption. Credentials are encrypted at rest with AES-256-GCM envelope encryption, bound to the record that holds them so an envelope cannot be replayed into another record. Data in transit is protected with TLS.
- Least privilege. An agent can perform only the typed actions you granted it; anything outside that grant is refused, as is any action not present in SafePod's catalogue.
- Content stays out of the audit trail. Connector activity records a hash of the input, not the input.
No service can guarantee it will never be breached, and we do not make that claim. SafePod is in early access and has not completed an independent third-party security certification. If that matters for your use, please treat it as the current state rather than a finished posture. We will notify affected users and, where required, the competent supervisory authority, without undue delay after becoming aware of a personal data breach.
10. Data retention and deletion
We retain personal data for as long as your account is active, and then as described below.
10.1 Disconnecting a connected service
Removing a connection deletes SafePod's stored credential for it, and your agents immediately lose the ability to act on it. This does not revoke the grant at the provider. SafePod does not call Google's token revocation endpoint, so to withdraw the permission at Google as well, remove SafePod at myaccount.google.com/permissions. Doing both is the complete removal.
10.2 Deleting an agent or a workspace
Deleting an agent releases its workspace storage and removes its conversation history, including the encrypted archive copy. Deleted workspaces are held briefly before permanent removal so that an accidental deletion can be reversed; after that window the data is gone.
10.3 Deleting your account
You can delete your account from SafePod's settings after re-authenticating. Deletion removes your account record, your conversation history, your stored model-provider credentials, your model egress configuration and your connector connections. Server logs age out on their own schedule and are not retained for marketing or profiling. We may retain limited records where necessary to comply with a legal obligation or to resolve a dispute.
11. Your rights and choices
Subject to applicable law, you have the right to access the personal data we hold about you; to have inaccurate data corrected; to have your data erased; to restrict or object to certain processing; to receive your data in a portable format; and to withdraw consent at any time without affecting processing already carried out. If you are in the EEA, the UK or Switzerland, you also have the right to lodge a complaint with your local supervisory authority.
If you are a California resident, you have the right to know what personal information we collect, use and disclose; to request deletion or correction; and to be free from discrimination for exercising those rights. SafePod does not sell personal information and does not share it for cross-context behavioural advertising, so there is no "Do Not Sell or Share" action to take.
In addition, within the product you can:
- use SafePod without connecting any third-party service at all;
- grant each agent only the actions it needs, and change or withdraw those grants at any time;
- delete an agent, a workspace, or your whole account.
To exercise any of these rights, contact [email protected]. We will respond within the period required by applicable law, and we may need to verify your identity before acting on a request.
12. Children's privacy
The Service is not directed to children and is not intended for anyone under 16. We do not knowingly collect personal data from children. If you believe a child has created an account, contact us at [email protected] and we will delete it.
13. Changes to this policy
We may update this policy from time to time. If a change materially affects how your personal data is used, we will update the date at the top of this page and notify account holders by email before the change takes effect. Your continued use of the Service after that point means the updated policy applies.
14. How to contact us
SafePod — [email protected]