Anthropic is pushing a draft bill to regulate AI models, but its safety thresholds face criticism. The bill lets the government block a model only if it poses a "foreseeable, imminent risk of more than $1 billion in damages." Many worry this narrow scope ignores other serious AI-related harms today.
As the push for AI regulation gathers pace, AI giant Anthropic has stepped up its efforts to shape a draft bill governing the industry, according to a report by Politico. The proposed legislation would allow the government to block the release of AI models it believes pose a danger to the public. However, the thresholds Anthropic has proposed for defining such risks have drawn fresh criticism over how the AI giant believes potentially dangerous models should be regulated.
As per the draft, “The bill would let the government block a model’s release or use only if it poses a foreseeable, imminent risk of more than $1 billion in damages, or of killing or seriously injuring more than 50 people,” Politico’s Owen Dahlkamp reported. This means that if a model were likely to kill or seriously injure fewer than 50 people, or cause less than $1 billion in property damage, the government would not be able to intervene under the proposed threshold.
What qualifies as a ‘risk’?
The bill further limits the circumstances that would justify intervention to three categories: the creation of a biological weapon, an AI model evading its developer’s controls, or a cyberattack. This means other forms of AI-related harm might not fall within the scope of the proposed legislation.
The bill would also prevent any US agency other than the Department of Commerce from intervening to suspend an AI model’s development or internal use.
Lawmakers from both parties have increasingly called for stricter AI regulation, proposing measures such as higher taxes on AI companies, pauses on frontier model development and a cabinet-level department to oversee the industry. Anthropic has stepped up its lobbying efforts, spending more in Congress during the first half of 2026 than it did throughout 2025.
It has also lobbied at the state level, including in California, where it spent at least $90,000 over the past year supporting policies favouring data centre development. The company has also engaged lobbyists in states such as Tennessee and North Carolina and reportedly sent a delegation to the Vatican to discuss AI consciousness with the Pope.
Anthropic’s proposed legislation seeks to centralise regulatory authority within a single federal agency. If passed, it could also limit states’ ability to introduce stricter AI regulations.