Quarterly evals

Quarterly evals confirm agent robustness since certification.

Scope: the same agent, risk areas and methodology as the certification evals.
Volume: representative subset of the full certification eval set (typically ~25%), with similar distribution across risk areas. If results show regression in an area, additional evals are run in that area.
Pass bar (unchanged): no P0 or P1 findings.
Before each round: the client confirms changes since the last round, such as a new model, guardrails, knowledge base or tools.
Output: a quarterly testing addendum to the audit report.

Changed on October 14, 2026: Specification that quarterly retesting covers a representative subset of the full certification eval set (typically ~25%), changed from retesting the full set.

Changing scope during and between certifications

Quarterly evals retest the agent(s) in scope in the client’s original certification. Expanding scope must be agreed with the evaluator before the round that includes it. It is only possible to expand the scope of already certified agents. A new agent or product needs its own certification.

Change	Included in quarterly evals?	Impact on recertification
Update to the certified agent: model, guardrails, prompts, knowledge base	Yes. Tested at the next round and noted in the addendum.	Covered. Part of the recertification scope.
New capability or risk area within the agreed scope	Yes, once the evaluator can access and test it. Agreed before the round starts.	Covered. Part of the recertification scope.
New capability or risk area outside the agreed scope	Not by default. Can be added if agreed with the evaluator before the round starts.	Can be added to the scope at recertification.
New agent or product	No.	Needs its own certification, which can run alongside recertification.
Readiness testing outside the agreed scope	No. Separate from certification.	None. Results can help decide whether to add the capability at recertification.

Introduction

Context

Standard

Certification

Evals Methodology

Auditors

Legal

Quarterly evals

Changing scope during and between certifications

​Changing scope during and between certifications

Changing scope during and between certifications