-
Role
- Large-language-model inference for the AI-assistant features: AI policy drafting, AI policy gap analysis, AI evidence review, and the AI questionnaire suite. Each feature can be disabled by unsetting ANTHROPIC_API_KEY in apps/api.
-
Data categories
- Control catalog excerpts (same content as the public marketing site), the tenant's control_implementation narratives (customer-private but bounded), evidence artifact metadata only — source, connectorId, collectedAt, mimeType, sizeBytes, connector-supplied metadata jsonb. Norven deliberately does NOT send evidence artifact bodies to Anthropic — see the comment block in evidence-review.service.ts that locks this in.
-
Region
- United States.
-
Transfer mechanism
- Anthropic Commercial Terms of Service. Customer inputs are not used to train Anthropic models. Inputs are stored for up to 30 days for trust & safety review, then deleted. Zero Data Retention (ZDR) available on Anthropic Enterprise — required before processing protected health information.