ARL is a vendor-neutral readiness yardstick; adopters answer to NIST AI
RMF 1.0 and the EU AI Act. Add arl_core::crosswalk mapping ARL's measured
axes to the RMF subcategories (and AI-Act articles) they evidence, plus
evidenced_by_claim(claim) returning exactly the subcategories a given
Claim supplies evidence for.
8 NIST RMF mappings (MAP 2.3, MEASURE 1.1/2.3/2.5/2.7/2.12/4.2, GOVERN
4.1) gated on the claim's axes/flags: validation depth, security class,
energy disclosure, methodology pre-registration, error-bars/failure-modes
publication. The standout is MEASURE 2.12 (environmental impact /
sustainability) <- ARL's energy axis, the dimension RMF/AILuminate barely
name. EU AI Act table (Art. 15 accuracy/robustness/cybersecurity, Art. 40
sustainability, Art. 53 GPAI docs).
Honest scope: an ARL claim EVIDENCES these subcategories, it is not a
compliance certification. Serializable tables; no new deps. 3 tests green
(populated+serializable, bare claim evidences nothing, strong claim
evidences energy/security/methodology), clippy clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
41eca62633