The obligations that actually reach an organization are audit findings, examination questions and customer reviews, and there are more of them every quarter. None of them accept a completion rate as an answer.
See what each one asks for or read the methodologyThe instinct is to start with the statute, because a statute sounds like the thing with teeth. In practice the statute is the weakest item on this list and the voluntary standard is the strongest, because the standard is the one with an assessor attached.
| Instrument | Status | What it asks for | Asks for evidence? |
|---|---|---|---|
| ISO/IEC 42001 clause 7.2 |
Voluntary, certifiable | Determine the competence needed by people whose work affects AI performance, ensure they have it, and retain documented information as evidence of it. | Yes, explicitly |
| NIST AI RMF GOVERN 3.2 |
Voluntary framework | Proficiency standards for the people carrying out system operation and oversight tasks. | Implied |
| FINRA Rule 3110 |
Binding, sector | AI training sits inside the supervision obligation. This is the one US requirement with examination behind it. | Yes, on examination |
| EU AI Act Article 4, as amended |
Binding, EU | Take measures to support the development of AI literacy. Since the July 2026 amendment it expressly does not require you to guarantee any level of competence in any individual. | Records suffice |
| US Dept of Labor AI Literacy Framework |
Voluntary guidance | Five foundational content areas. Aimed at the public workforce system rather than at employers directly. | No |
| US state law CO, IL, CA, TX, CT, NYC |
Binding, varies | Notice, bias audit and anti-discrimination duties around automated decisions. | No training duty |
ISO/IEC 42001 is the management system standard for AI, and it is certifiable. That single property is what separates it from everything else on this page. A voluntary framework asks you to consider something. A certifiable standard sends a person to your office who asks to see it.
Clause 7.2 asks an organization to determine the competence necessary for people whose work affects AI performance, to ensure those people are competent, and to retain documented information as evidence of that competence. Clause 7.3 covers awareness of the AI policy and of what happens when it is not followed.
An assessor holding that clause and looking at a stack of course completion certificates is looking at evidence of attendance. The gap between what the clause asks for and what most organizations can produce is the entire reason this company exists.
The sequence is consistent. AI tools go in. A policy follows, usually quickly. Then someone above the program asks how anyone knows the policy is being followed, and the honest answer is that nobody does.
Article 4 has applied since 2 February 2025 and reaches organizations with no European entity, through customers, staff or output placed in the Union. The Digital Omnibus rewrote it in July 2026, converting a duty to ensure a sufficient level of AI literacy into a duty to take measures supporting its development, and adding an express statement that no specific level in any individual has to be guaranteed. The standard got softer. Because there is no longer an outcome to point at, the record of what you did became the only assessable thing.
We keep a longer treatment of the obligation, the timeline and who is in scope, and a scoping tool that produces a dated determination you keep.
Every instrument on this page is easier to answer with a measured population than with an attendance record. That is the whole argument.