Obiguard gives Risk and Legal the immutable, framework-mapped evidence trail they need — without waiting on engineering to pull logs before every audit.
Coverage is reported against NIST AI RMF, ISO/IEC 42001, the OWASP LLM Top 10, Malaysia’s AISCF, Singapore’s MGF for GenAI, and your internal AUP — scored from your own platform activity, category by category.
Every event is append-only and cryptographically signed. No editing, no deleting. The record you hand the auditor is the record Obiguard wrote at the time of the event.
Pull a filtered export for any audit request — by framework, by time range, by agent, by control. Download as CSV, JSON, or stream to your GRC tool.
Edge cases and high-risk violations route to the Review Queue. Every review decision — approve, escalate, annotate — is written back to the ledger with the reviewer identity.
The Ethics & Bias report scores a random sample of each day’s traces against six judge criteria — toxicity, demographic bias, sentiment consistency, fairness, misinformation, privacy leakage. Evidence that trustworthy characteristics are evaluated, not just asserted.
Every Project Moonshot evaluation run is retained against the project — target, prompt under test, grade, and failing recipes — which is what a red-teaming control asks you to produce.
Toxicity, bias, sentiment consistency, fairness, misinformation, privacy leakage.
NIST AI RMF, ISO/IEC 42001, OWASP LLM Top 10, AISCF, and MGF for GenAI — as a self-assessment aid.
Evidence is written at event time — no after-the-fact assembly.
From first call to enforced policy with full audit trail.