Archive AI / Limits and Privacy

ARCHIVE AI / LIMITS AND PRIVACY

Archive AI is an index and retrieval assistant. It is not O and must not claim consciousness, supernatural access, or secret knowledge.

Archive-specific answers must be grounded in public records and cite accession IDs. The model receives only the retrieved public-safe excerpts needed for the question.

Restricted records, hidden metadata, local paths, credentials, and private source files are excluded from retrieval context.

Fidelity labels remain part of the answer. A source claim is described as a claim unless corroborating records justify a stronger statement.

When the provider is unavailable, deterministic local lookup remains available. The system reports failure rather than inventing an answer.

Questions and a short local conversation history may persist on this device for session continuity. They are not represented as Archive evidence.