06 / REGULATORY INTERFACE

EU AI Act Mapping

A bounded correspondence between selected EU AI Act obligations and the evidence, provenance, oversight and runtime questions used in the ARAYUN_173 research architecture.

What the mapping does

It identifies where selected regulatory requirements create questions that can be examined through bounded evidence: what was decided, which sources support the claim, who held authority, what runtime path executed, and whether a named control was actually bound to that path.

Selected area	Regulatory focus	ARAYUN_173 inspection question	Claim boundary
Article 9	Risk management	Are risks, evidence gaps and intervention points documented across the named lifecycle?	Mapping does not prove risk control effectiveness.
Article 13	Transparency and information	Can the decision, sources, transformations and limitations be traced for the specified claim?	Traceability does not establish correctness.
Article 14	Human oversight	Is accountable intervention authority defined, reachable and evidenced?	An interface or policy does not prove effective oversight.
Article 15	Accuracy, robustness and cybersecurity	Which metrics, tests and runtime evidence support the bounded performance claim?	One metric or local PASS does not establish system-wide stability.
Article 50	Transparency obligations	Which disclosures apply to the named system, output and deployment context?	The mapping is not a legal determination of applicability.

---

Evidence before conclusion

Documentation, runtime evidence and enforcement are separate states. A documented control may remain not evidenced in execution. A verified local enforcement path may support only the named effect and cannot be generalized to repo-wide, system-wide or OS-wide control.

Commercial decision boundary

The Executive AI Decision Review can examine one submitted decision against its available evidence and authority boundary. Its report is a bounded review, not certification, legal opinion or formal compliance audit.