Trusted AI infrastructure, built on transparency.

Zero model training on customer data

Nothing routed through Founder OS is used to train any AI model that benefits another customer.

"No user-submitted data is used to train any language models, nor is it shared externally or made viewable by any third party."

Single-user authentication

Founder OS is a personal cockpit. One password (set via FOUNDEROS_PASSWORD env var) gates the entire app. Sessions are HMAC-signed cookies, HttpOnly, SameSite=Lax, Secure-in-production. CSRF tokens validate every form submission. Five failed login attempts from an IP triggers a 15-minute lockout.

Encrypted in transit and at rest

TLS for every connection (Let's Encrypt). At-rest encryption on the hosting volume. Role-based access controls and MFA on every infrastructure surface. API keys (Anthropic, Brave) live in environment variables, never on disk.

Atomic writes

Every data file write — agents.json, pipeline.json, per-agent KPIs, drafts, send logs, audit log — uses a write-temp-fsync-rename pattern. A crash mid-write leaves the old file intact, never a half-written target.

Audit log

Security-relevant actions (sign-in, sign-in failure, lockout, sign-out, pipeline mutations, agent run, agent delete) are written to a separate append-only NDJSON file at data/audit.log. Passwords, API tokens and request bodies are never logged.

72-hour breach notification

If a data-protection incident is detected, affected parties are notified within 72 hours of discovery with the minimum context required to exercise their obligations under UK GDPR Article 33.