Data governance

Retention and provider policy

Last updated August 20, 2026

Data classes

Raw provider responses

Raw response storage is controlled per provider. When permitted, payloads have a retention deadline and may move to private object storage. Expiration removes payload bytes while preserving an immutable hash and redaction record so historical metrics remain explainable.

Provider boundaries

OpenAI, Anthropic, and Perplexity require separate credentials, current policy approval, and compatible retention settings. Google Gemini search grounding remains disabled because its current terms are not compatible with cross-run stored analytics without written authorization.

Exports

Company data exports omit passwords, sessions, secret references, integration secrets, and raw provider payloads. Report exports neutralize spreadsheet formulas and use private, non-cacheable downloads.

Deletion workflow

A workspace owner must re-enter the company name and password. Pigeon Radar records the request, schedules a seven-day cooling-off period, and allows cancellation. Final production erasure is an operator-controlled workflow so immutable evidence, billing duties, legal holds, and backups can be handled safely.

Website audits

Audits require an authorization attestation. The crawler respects robots.txt, blocks private and reserved networks, validates redirects, limits response size and time, and stores sanitized evidence rather than executable page content.