What happened
Anthropic is privately circulating a draft legislative proposal on Capitol Hill that seeks to establish a federal framework for regulating frontier AI models. The draft, as reported by Politico, proposes that the federal government only be permitted to block or suspend an AI model if it poses an imminent, foreseeable risk of causing more than $1 billion in damages or resulting in more than 50 deaths or severe injuries. The proposal mandates that such interventions require a federal judge's order rather than unilateral action by executive agencies.
The draft legislation, which has not been previously reported, outlines a system where independent evaluators registered with the Commerce Department would conduct periodic assessments of frontier AI laboratories. These evaluators would be required to report any 'imminent enumerated risk' to the Commerce Department within 72 hours.
Under the proposed bill, the Commerce Department would be stripped of the power to unilaterally shut down a model. Instead, the department would be required to petition a federal judge in the District Court for the District of Columbia to obtain an order for suspension.
The bill explicitly limits the scope of federal intervention to risks arising from cyberattacks, the creation of biological weapons, or instances where a model evades its developer's controls. It also includes provisions for mandatory safety frameworks, transparency reports, and whistleblower protections.
The proposal seeks to override state and local regulations regarding public risk disclosures and third-party audits, a move that aligns with industry arguments that fragmented state laws create excessive compliance burdens.
Source details: businesstoday.in ↗
Why it matters
This proposal represents a significant attempt by a major AI developer to shape the regulatory environment governing its own industry. By advocating for a high threshold for government intervention and the pre-emption of state-level regulations, Anthropic is seeking to standardize compliance requirements while limiting the potential for arbitrary or rapid federal shutdowns. The move highlights the ongoing tension between industry desires for regulatory clarity and the government's need to manage risks associated with frontier AI, such as cyber-espionage or the creation of biological weapons. The proposal's emphasis on judicial oversight and federal pre-emption directly challenges current legislative trends in states like California, which have pursued more stringent, localized safety mandates.
The proposal follows recent, non-public actions by federal officials who placed export controls on two of Anthropic's advanced models. This experience appears to have motivated the company to seek a more predictable, legally defined process for government intervention.
By setting the intervention threshold at $1 billion in damages or 50 casualties, the bill attempts to define 'catastrophic' risk in concrete terms, potentially shielding developers from regulatory action based on smaller-scale safety incidents or routine probes.
The debate over federal pre-emption is a major political hurdle. Democrats in Congress have generally resisted efforts to weaken state-level safeguards, meaning the inclusion of this provision may make the bill a difficult sell in the current legislative climate.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
crm_get_transaction(id='4092').Why can ethical evaluation not be reduced to one model score?
What to watch next
The primary point of contention will be the proposal's federal pre-emption clause, which seeks to override state-level safety regulations. This is expected to face significant resistance from lawmakers who favor stronger, localized protections. Additionally, the effectiveness of the proposed independent evaluation system—where companies pay auditors directly—will likely be scrutinized for potential conflicts of interest. Observers should monitor whether this draft gains traction among federal legislators or if it serves primarily as a lobbying tool to influence the broader debate on .
Watch for reactions from state-level regulators and lawmakers in jurisdictions like California, who may view the federal pre-emption clause as an attempt to undermine their authority to protect citizens from AI-related harms.
Monitor whether other major AI labs align with Anthropic's proposal or if they advocate for different regulatory structures, as the industry remains divided on the appropriate balance between safety and innovation.
The role of the Commerce Department in this proposed framework is critical; the bill would fundamentally change the agency's relationship with AI developers by shifting the burden of proof for safety interventions to the judicial system.