Verdicts
Six verdicts, one evidence standard
Verdicts are technical guidance, not certification or legal approval. Every verdict is reached through peer review and can be revisited.
Evidence supports production use without additional guardrails for common use cases.
When it's used: When sandbox testing and review found no unresolved high-severity risks.
Production-viable, but only with specific mitigations in place.
When it's used: When known risks have practical, documented guardrails.
Suitable for internal or lower-stakes use, not yet validated for customer-facing production.
When it's used: When risk tolerance for the evaluated use case is lower than typical production.
Promising, but only validated in a sandbox context so far.
When it's used: When testing is still early or evidence is incomplete.
Known, unresolved risks make production use inadvisable at this time.
When it's used: When testing found high-severity issues without a practical guardrail.
No verdict has been reached. Treat any draft content as non-authoritative.
When it's used: Default for every report until it is formally published.