Anthropic drafts legislative proposal with strict limits on federal AI blocks

Artificial intelligence lab Anthropic has been shopping around a draft bill on AI safety for consideration by the United States Congress, TechCrunch AI and Truthout reported, citing an earlier report by Politico. TechCrunch AI and Futurism reported that the proposed legislation establishes specific scenarios under which the government can block an artificial intelligence model from distribution. Specifically, TechCrunch AI and Truthout reported that the proposed framework prevents federal agencies from halting a model release without approval from a judge. [1][4][5]

Under the draft bill, federal regulation and intervention are limited to severe threshold events. TechCrunch AI, UA.NEWS, Truthout, Futurism, and Cybernews all reported that the draft legislation establishes a death threshold where a model could be blocked if it poses a risk of killing more than 50 people. As FTMQ SI reported earlier, the draft legislation also requires 1 billion dollars in damage before regulators can step in, alongside provisions for state preemption and expanded transparency mandates. [1][3][4][5][6][9]

The legislative push comes amid broader political and public debates surrounding artificial intelligence safety and governance. Concerns over ethics, alignment, misinformation, and technological safety have expanded significantly during the recent artificial intelligence boom. Anthropic operates under an independent trust model and uses Constitutional AI, a training method that aligns model behavior using natural-language principles rather than relying solely on human preference feedback. [7][8]

The draft proposal represents an effort by Anthropic to shape how the United States government might regulate the artificial intelligence industry. Debates continue among lawmakers and industry stakeholders over the proper framework for balancing technological safety against regulatory restriction. [1]