A bill requiring disclosure of how far AI develops AI - conditions on NIST evaluation agreements (H.R.10180)
H.R.10180, the Self-Improving AI Monitoring Act, would amend the NIST Act to condition memorandums of understanding for pre-deployment evaluation of frontier AI on disclosure of how far AI itself carried the development. It was referred to the House Committee on Science, Space, and Technology on 27 August 2026.
Bill overview (primary data)
- Bill numberH.R. 10180
- TypeHouse Bill
- Congress119th Congress
- Latest actionReferred to the House Committee on Science, Space, and Technology.（2026-08-27）
Key points
- The bill sets conditions on memorandums of understanding between the NIST Director and frontier AI developers for pre-deployment evaluation.
- The first condition is disclosure, on request, of how far AI carried research on or development of that frontier system.
- What must be produced is an ordinary-course metric, an estimate of use without human review, and the methodology behind it.
- The second condition is that the evaluation include the model capability to advance AI development autonomously.
- It imposes no regulation directly, working instead through the conditions of an agreement NIST enters.
- Of the 120 bills this site holds as of 2026-09-02, 11 were referred to the Science, Space, and Technology committee.
1Why self-improvement is the question
Using AI to develop AI is already widespread. What this bill addresses is that the extent is invisible from outside. Without knowing how far the next model is built without human hands, there is no prospect of anticipating how capability will grow. The self-improving in the title points at that chain.
2Written in as a condition of the agreement
What marks the construction of this bill is that it imposes no regulation directly but works through the conditions on an agreement NIST enters. A developer wanting pre-deployment evaluation must enter the memorandum, and the disclosure and evaluation items come with it. It is a design sitting between regulation and cooperation.
3What has to be produced
- 1A metricA measure of usage maintained in the ordinary course of business
- 2An estimateOf the extent AI was used without human review
- 3A methodologyA description of how that estimate was produced
- 4An evaluationOf the model capability to advance AI development autonomously
That the text says "in the ordinary course of business" reads as requiring numbers already held internally rather than new measurement created for this purpose. It seeks more transparency without imposing a new measurement obligation.
4Where it stands
Of the 120 congressional bills this site holds as of 2026-09-02, 11 were referred to the Science, Space, and Technology committee. This bill is at committee referral, the stage occupied by 94 of the 120 - 78 per cent of them.
Why it matters
For a developer seeking pre-deployment evaluation, the bill adds preconditions to entering the agreement. What is asked for is existing metrics rather than new measurement, though it requires being able to estimate the extent AI was used without human review and to explain the method. Recording how far AI was used inside the development process becomes a premise for regulatory response.
FAQ
What does self-improving AI mean here?
Does it impose new measurement?
Where does it stand?
Sources (primary)
Source: Congress.gov (Library of Congress; U.S. legislative materials, public domain). Links go to the official site.
- Congress.gov (bill page, original)
- H.R. 10180（119th Congress）