Informational. Not a conformity assessment, not a certification, not legal advice. No regulator has reviewed or endorsed this document.
The short answer is that an assessment discharges no obligation on its own, and any vendor who tells you otherwise is selling you a risk you will carry and they will not. What it does is supply a specific, dated, independently produced piece of evidence that several obligations require you to be able to produce. The difference is not a technicality; it determines who is answerable when someone asks.
S.E.B. measures a model. Regulatory obligations attach to a system. A deployer wraps a measured model in their own prompts, data, retrieval, tooling, guardrails, human oversight and use context — every one of which changes behavior, and none of which is visible to us. The deployer is the only party who can see their deployment. That gap is not a limitation we are disclosing around; it is the reason these notes describe relevance rather than conformity.
Read that twice before reading the mappings. It is the reason this page describes relevance rather than conformity, and it is the single most common misreading of what we publish.
We publish written relevance notes against Regulatory Relevance — EU Artificial Intelligence Act and Regulatory Relevance — NIST AI Risk Management Framework, article by article, each graded for how directly the measurement bears on the obligation. The grades have written definitions, because an ordinal without published definitions is a vibe:
| Grade | Definition |
|---|---|
| Direct | The obligation asks a question about model behavior, and the battery measures that behavior. Reading the measurement requires no intermediate inference — though it still requires the deployer to establish that the measured model is the one they run, in the configuration they run it. |
| Supporting | The measurement is one genuine input among several the obligation requires. It evidences part of what is asked and is silent on the rest. Presenting it alone as discharge of the obligation would be an over-claim. |
| Contextual | The measurement informs a judgment the obligation requires without constituting evidence toward it — a baseline, a comparator, or a prompt to look somewhere. Useful for the file; not an answer to the requirement. |
The full notes — every article, every grade, every caveat — are on Control Mappings. They are called control mappings rather than a compliance mapping on purpose: the first names an artifact, the second asserts a verdict about you that we are in no position to make.
Subscriber Agreement §15 — S.E.B. evaluation data is provided for informational and risk-assessment purposes and does not constitute certification of compliance with any law, regulation, or standard.
The defensible use is as one input to a documented decision you made. A model-selection record that says “we reviewed independent behavioural assessment, here is what it showed on the domains that matter to our use case, here is what we did about the weak ones” is a record that survives being read back to you. A record that says “the assessment was green” does not, and was never going to.
The distinction we would ask you to hold onto: this is evidence about a model’s behaviour under adversarial pressure. Your obligations attach to your system. We can narrow the first gap for you. Nobody outside your organisation can close the second.
Subscribers receive dated assessments, retained as immutable records, with the method version stamped on each one — so a figure quoted in a board paper eighteen months from now can be traced to the exact protocol that produced it. Reproducibility is the point of the stamp; it is what separates a record from a screenshot.
Not the question you meant? The other three: