A four-page House bill text, captured from Representative Pat Harrigan's House forms subsite, amending section 22A of the National Institute of Standards and Technology Act (15 U.S.C. 278h–1) — the section that carries the authorities exercised by the Center for AI Standards and Innovation. Section 1 gives the short title as the "Self-Improving AI Monitoring Act."
The document is an unnumbered discussion draft. Its face reads "H. R. ll" with a blank committee referral line and a blank original-signature line, so no bill number attaches to this text. Its introduction line names only "Mr. WHITESIDES" — Representative George Whitesides (D-Calif.). Trade coverage on September 4, 2026 attributed the measure to Harrigan and Whitesides jointly and described it as introduced on August 31, 2026 (Source: nextgov.com); the document itself records neither Harrigan nor an introduction.
The two mechanisms
A new assessment duty. Section 22A(b) is amended to add a paragraph (6) directing the Director to "assess, based on the information, if any, disclosed pursuant to subsection (h)(1) and other information, trends regarding the capabilities of artificial intelligence systems, including such a capability to autonomously facilitate or conduct research on or development of an artificial intelligence system, that might contribute to scientific research and development breakthroughs." The subject of the assessment is the trend across systems rather than any one system, and the input is explicitly the disclosures obtained under the second mechanism.
A prohibition on the Director's own contracting. A new subsection (h) provides that in carrying out subsection (b), the Director "may not enter into with the developer of a frontier artificial intelligence system (as determined by the Director) a memorandum of understanding to conduct pre-deployment evaluation of such frontier artificial intelligence system unless" two requirements are satisfied:
- The memorandum requires the developer, in carrying out the evaluation, to disclose to the Director on request "information relating to the extent to which an artificial intelligence system was utilized to facilitate or conduct research on or development of such frontier artificial intelligence system." Three examples are given: a metric the developer maintains in the ordinary course of business relating to that use; an estimate of the extent of such use without human review; and a description of the methodology behind the estimate.
- The evaluation itself includes "an evaluation of the capabilities, if any, of such frontier artificial intelligence system to autonomously facilitate or conduct research on or development of an artificial intelligence system."
The obligation runs against the Director rather than against developers. A frontier developer is bound only through a memorandum of understanding it chooses to enter, and the definition of a frontier developer is left to the Director. The drafting is a constraint on what NIST may agree to, so its practical reach depends on how many developers seek pre-deployment evaluation at all.
The first example — "a metric maintained by such developer in the ordinary course of business" — asks for a figure a company already keeps rather than one produced for the government, which limits the compliance burden and correspondingly limits what can be demanded of a developer that keeps no such metric.
Relation to the surrounding legislation
The draft addresses the same object as the frontier-evaluation provisions of the Obernolte–Trahan discussion draft (Frontier Act / Great American AI Act (Obernolte–Trahan)) but through a different lever: where that draft builds an evaluation regime, this one conditions NIST's authority to enter into the agreements through which evaluation happens. The capability it names — a system's ability to autonomously conduct AI research and development — is the object of Recursive Self-Improvement (RSI) and Automated Alignment Research.
Relationships
- related: NIST CAISI (Center for AI Standards and Innovation) — the amended section carries its authorities.
- related: Recursive Self-Improvement (RSI); Automated Alignment Research — the capability the bill directs NIST to assess.
- related: AI Pre-Release Vetting; Independent Verification Organizations (IVOs) — the evaluation regime the memorandums sit within.
- related: Frontier Act / Great American AI Act (Obernolte–Trahan) — the other live federal frontier-evaluation instrument.
- related: Self-Improving AI Monitoring Act — the tracking page for the measure.