Brief
Anthropic drafts AI safety bill for Congress, source says
The startup’s proposal mirrors California’s SB 53 and sets a $1 billion damage or 50‑person injury trigger for federal action.
By Felo News Desk · Published
Anthropic has been privately sharing a draft bill that would regulate frontier AI safety and security with U.S. lawmakers, according to a report by POLITICO cited by Mashable.
What happened
The confidential draft, obtained by POLITICO, outlines criteria for federal officials to halt a model’s launch 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." The bill also requires the danger to stem from a bioweapon, loss of developer control, or a cyber‑attack. It designates the Commerce Department as the sole agency that can seek to stop a model, creates outside evaluators, and would override certain state laws.
What the reports add
Mashable notes that the language closely follows California’s SB 53 and that Anthropic has previously supported two California AI safety laws. The draft also proposes new duties for AI labs and a mechanism to supersede state regulations.
What was said
An Anthropic spokesperson told POLITICO the bill reflects the company’s Advanced AI Framework and called its proposals "sensible policies" for Congress.
How it came about
Anthropic’s lobbying follows its recent push for industry‑wide safety standards, including CEO Dario Amodei’s call to "pace the frontier" and the White House Accord signed by leading AI CEOs. Earlier this month, Felo reported on FTC subpoenas to Anthropic and OpenAI over AI‑related safety incidents (ftc-subpoenas-anthropic-openai).
Key facts
- Anthropic is privately pitching a draft AI safety bill to Congress (mashable.com)
- The bill would allow federal action only if a model poses over $1 billion in damages or harms more than 50 people (mashable.com)
- The draft mirrors California’s SB 53 and designates the Commerce Department as the sole halting authority (mashable.com)
- Anthropic’s spokesperson described the proposals as "sensible policies" (mashable.com)
Sources
- [1] mashable.com — originally reported as “Anthropic is pitching Congress its own AI safety bill, report says”
