Privacy

Evidence remains
under human control.

Original evidence stays in the governed archive. External analytical tools receive only explicitly authorized, verified working derivatives.

AI processing boundary

Minimal, transient, accountable.

Independent Google Gemini and Anthropic Claude analysis uses bounded requests. Google uses the GBAuthentic-controlled Vertex AI project and the United States multi-region endpoint. Project-wide Gemini caching is disabled. OpenAI is deferred and inactive.

01

No originals

Customer photos, forensic originals, completed reports, credentials, signed URLs, and private storage paths are excluded from every active provider request.

02

No provider persistence

The application disables provider storage and caching where supported and does not create provider files, stored conversations, cached-content resources, resumable sessions, grounding requests, or provider-hosted retrieval resources.

03

Governed derivatives

Only an authorized, hash-verified working derivative and the bounded analytical instructions may be sent for an active request. Each active provider receives the same evidence independently and never receives another provider's output.

04

Human accountability

Provider output is advisory, attributable, and subject to qualified human review. Provider agreement cannot determine authenticity or authorize publication.