Security & Data Handling
Space Office is built around human-reviewed AI work running inside your own workspace, using your own AI provider keys. This page explains the controls in place today and the standards we keep tightening as we grow.
Last updated: June 18, 2026
No unsupported compliance claims
We are not claiming SOC 2, HIPAA, ISO, or similar certification on this page. When we complete a formal program, we will publish the details plainly rather than imply them.
Encryption and credential storage
AI provider keys and OAuth tokens are encrypted at rest using AES-256-GCM before they are written to our database.
The encryption key lives in our backend environment and is never shipped to the browser. Encrypted credentials are never returned to the client — the dashboard only ever sees connection status, not the secret itself.
Please do not send passwords, private keys, recovery codes, or customer credentials through the contact form or email. Add them only through the product's secure connection flows.
Bring-your-own-key model
You connect your own Anthropic or OpenAI account using an API key, a device-code flow, or an OAuth popup. Your prompts and content are processed under your own provider account and that provider's terms.
Provider credentials are scoped to your workspace and used only to run your work. Space Office does not pool customer content to train its own models.
Connected apps and least privilege
App connections such as Google, Notion, Figma, and X use OAuth with PKCE and one-time, short-lived state values to protect the connection flow.
Agents use connected access only to perform the work you asked for, and you can disconnect an app at any time, which removes the stored credential for that connection.
You can also bring your own OAuth application credentials; if you do, the client secret is stored encrypted just like other secrets.
Workspace isolation and access
Customer data is scoped to your workspace, and queries are constrained to the workspace that owns the data.
Access to operational systems is limited to the people and systems that need it to provide, support, or secure the service.
We continue to tighten authentication, permissions, auditability, and retention as the product matures.
Human-in-the-loop review
Hydrogen plans each project, assigns the right specialists, and reviews their outputs before anything is presented as done.
Plans wait for your approval before work begins, and deliverables sit in a review state until you mark them complete. That double gate is how Space Office turns 'AI made something' into 'AI made something good.'
Responsible disclosure
If you believe you found a security issue, email hello@spaceoffice.ai with "Security" in the subject line, a clear description, and steps to reproduce.
Please do not access, change, delete, or share data that does not belong to you while testing.
We do not currently operate a paid bug bounty program, but we appreciate and act on good-faith reports.