Privacy Policy
Last updated: October 2026
1. Data Architecture & Isolation
Firebreak is engineered with zero-trust architectural isolation. Workspace tenancy is isolated at the database query layer, ensuring that no incident telemetry, repository metadata, or configuration is accessible across different organizational boundaries.
2. Encryption of Third-Party Credentials
GitHub installation tokens, Vercel API credentials, and user-provided AI provider API keys are encrypted at rest using AES-256-GCM. Cryptographic operations bind workspace-specific Additional Authenticated Data (AAD) into every ciphertext payload, preventing ciphertext transplant attacks across accounts.
3. AI Provider Usage & Data Sharing
When an AI analysis is requested for a suspect commit diff, code snippets are transmitted strictly to the LLM provider configured by the workspace administrator (e.g., Anthropic, OpenAI, Google Gemini, or xAI). We do not train proprietary models on your private source code.
4. Audit Logs & Data Retention
Incident events, webhook delivery receipts, and verification probe streak metrics are retained for auditability and compliance reporting. You may request cryptographic deletion or workspace evacuation at any time.