NIST AI Risk Management Framework (AI RMF 1.0) covers 25.6% of EU AI Act
11 of the 43 controls in EU AI Act are already satisfied by evidence you collected for NIST AI Risk Management Framework (AI RMF 1.0). 32 are genuine gaps. Every claim below was judged against both control sets and then argued against; the ones that did not survive are published further down with the reason each failed.
What this leaves you to do
EU AI Act has 43 controls. Holding NIST AI Risk Management Framework (AI RMF 1.0) already evidences 11 of them, so the work in front of you is 32 controls, not 43, which is 74% of the standard rather than all of it.
That is the whole claim. We do not know your hourly rate, how long a control takes you, or how many people you have, so there is no figure here in dollars or weeks. Every number in that sentence comes from the two counts above it and can be re-derived from the free tools without taking our word for any of it.
This number is directional. It says how much of EU AI Act your NIST AI Risk Management Framework (AI RMF 1.0) evidence satisfies. The reverse pair is a different number, often very different, because a security standard has enormous depth for access control and almost none for lawful basis or data subject rights.
41 candidate mappings were examined and 9 were removed. Signed off 2026-08-20, review level machine verified. Mappings were judged by Claude Code rather than read line by line by a practitioner. Every claim shows its reasoning so you can check it. Ask and a practitioner will review this pair.
Where the gaps are
Coverage is never evenly spread. A source standard usually satisfies one part of a target almost completely and barely touches another, and which part is which is the thing worth knowing before you plan the work.
Theme level, not control level, deliberately. The per-control list of what is evidenced and what is a gap is the report itself, so publishing it here would be publishing the thing being sold.
Claims that held
A sample. Each one names the control whose evidence does the work, the control it satisfies, and why.
Disaggregated fairness evaluation with stated definition satisfies the examination for possible biases.
Documented data collection and selection with representativeness and suitability is the data governance limb.
Validated explanation pitched at the acting audience gives operational transparency of the output.
Documented knowledge limits and output use, delivered to deployers, is the instructions for use duty.
Tested disengage and deactivate mechanism with assigned role is the intervene and stop limb.
Oversight process assessed for whether it is exercisable, with the overseer's authority recorded.
Safe-failure testing beyond knowledge limits with residual risk against tolerance is the robustness limb.
Pre-deployment validity and reliability demonstration with generalisability limits is the accuracy limb.
Claims that did not hold
9 proposed mappings for this pair were rejected. They are kept in the graph rather than deleted, so what was thrown out is as inspectable as what survived. A crosswalk that never rejects anything is not being judged.
Judged against a node bundling seven distinct provider obligations. The recorded intent says only provider obligations against roles and responsibilities, which sits equally on Art.16, the list of provider duties, and on Art.17(1)(m), the accountability framework inside the quality management system. Two candidate articles is not an unambiguous re-home.
Claimed at refuted confidence before it was rejected.
Refuted in review.
Refuted in review.
Refuted in review.
Refuted in review.
Refuted in review.
Refuted in review.
Claimed at unrated confidence before it was rejected.
Refuted in review.
The full report
Everything above is a sample. The report is every evidenced control and every gap, with the reasoning and the source document behind each one, in a form you can hand to an assessor. $299, emailed immediately.
Buy this crosswalk