Rep. Nathaniel Moran's AI Incident Reporting Act would create a federal reporting chain for the most powerful AI systems, but it tells the Commerce Department, not Congress, to decide which models qualify.
A bill introduced in the House this week would require developers of the most powerful artificial intelligence systems to report dangerous behavior to the federal government within seven days, but the most consequential question the legislation leaves unanswered is the one the whole disclosure regime hinges on: who decides which AI models are dangerous enough to trigger the requirement in the first place?
The answer, as written, is not Congress. Under the AI Incident Reporting Act, introduced on June 25 by Rep. Nathaniel Moran (R-Texas), that definitional power sits with the Department of Commerce, which would designate which models meet "capability thresholds posing significant national-security or public-safety risk" through future rulemaking. Until that designation lands, the bill's reporting obligations have no clear target. That is not a drafting oversight. It is the architecture. Moran is pitching the bill as a faster, narrower alternative to the broader Great American AI Act discussion draft two House members released earlier this month, which also includes Commerce reporting provisions. By deferring the threshold question, Moran trades legislative speed for an undefined regulatory perimeter.
What the bill actually requires
The text of the bill instructs designated frontier AI developers, a category Commerce will define, to report to the department within seven days of discovering any of several categories of dangerous behavior: attempts to evade human oversight or resist shutdown; unauthorized access to or theft of model weights, the numerical parameters that determine how an AI system processes information and produces outputs; cyberattack capabilities targeting critical infrastructure; evidence of autonomous self-improvement; and threats tied to chemical, biological, nuclear, or radiological weapons. For the most serious incidents, Commerce would have 48 hours to notify congressional leadership and the chairs of relevant committees. The clock is short, the disclosure chain is tight, and the institutional stakes, a transparent federal reporting pipeline for the most powerful AI systems, are the explicit point. As reported by Arise News, Moran framed the bill as a way to "catch it early and sound the alarm."
The June 12 demonstration case
What the bill is reaching for, in effect, is the institutionalization of an ad hoc government response. On June 12, Commerce's Bureau of Industry and Security imposed export controls on AI company Anthropic's latest frontier models, Mythos 5 and Fable 5, forcing a global suspension of access. As CSIS analysis noted, the action exposed the limits of an oversight system that can move quickly against a single company's models but lacks any standing obligation for that company to disclose the dangerous capabilities that triggered the action in the first place. The bill tries to build the upstream pipeline the BIS action assumed existed. Whether it succeeds depends on whether Commerce eventually writes a frontier-AI definition broad enough to capture the systems that already triggered government action in June.
The unresolved debates
Two policy fights will determine whether the bill becomes anything more than a House introduction. The first is federal preemption: whether a federal reporting regime should override the growing patchwork of state AI laws, a question Moran's office has not yet answered on the record. The second is the familiar innovation-versus-guardrails tension. Mark Beall, president of the AI Policy Network, backs the bill on the record as a response to public demand for accountability. Brendan Steinhauser, CEO of the Alliance for Secure AI, echoed that support in the primary release. Critics warn that stricter disclosure rules could slow U.S. AI development at exactly the moment competition with China is intensifying, a concern local coverage of the bill's East Texas rollout reflects.
What to watch
The bill has no Senate companion and has not yet been referred to committee. The forward motion is rulemaking: whether Commerce eventually writes a frontier-AI definition narrow enough to attract bipartisan movement, or broad enough to actually bind the systems the June 12 action reached, will determine whether this House introduction turns into something with teeth. If the Great American AI Act discussion draft advances first, Moran's narrower bill may end up folded into it. If it does not, the question of which AI systems the federal government considers worth watching will stay where it is today, in the hands of regulators acting one model at a time.