The safety warnings from Anthropic researchers that dominated AI discourse this week have now produced something concrete in Washington: a legislative push. Reuters reported Thursday that US lawmakers are calling for new AI rules in response to the researchers' statements, while Massachusetts Representative Lori Trahan used a Thursday morning interview to detail her Frontier Act, a bill that would impose binding obligations on frontier AI developers for the first time.
Speaking on WBUR's Morning Edition, Trahan said the reaction on Capitol Hill to this week's events has been unlike anything she has seen since she began working on AI regulation earlier this year. For daily coverage of AI policy moves in Washington, Brussels and beyond, follow AI Buzz Wire, your source for breaking AI news and regulation updates.
What the Frontier Act Would Do
Trahan, who co-authored the bipartisan bill, described it as a set of "rules of the road" for the handful of companies whose models could cause catastrophic harm. The Frontier Act would mandate four things, according to the congresswoman: transparency requirements for frontier AI developers, incident reporting, independent verification of model safety claims, and what she called a government pause button if a model poses an imminent risk of catastrophe.
"At a basic level, we need to put guardrails on the frontier labs," Trahan said, framing those provisions as "table stakes" while Congress works out policy for the many other sectors AI will touch.
'The Honor System': Trahan's Case for Binding Rules
The core of Trahan's argument is that nothing currently binds the developers of the most powerful AI systems. "Today, the only thing we have in place is the honor system," she said. "The most powerful technology in human history is being regulated by its developers."
She acknowledged skepticism among colleagues, saying some view the week's warnings as fear-mongering or too extreme. But she described a rapid shift in the mood of the chamber: "My phone hasn't stopped ringing since Jacob Coxon blew the whistle. I think my colleagues are scared. They're hearing from their constituents who want us to act... This week has just hit a tipping point."
Notably, she said the pressure is not confined to one party. "It's not just Democrats," Trahan said, noting that "MAGA Republicans called for bipartisan action yesterday. The ground has really shifted."
The Warnings That Started It
The legislative ferment traces directly to statements by two Anthropic researchers. Jacob Coxon quit the company on Tuesday, writing on social media that Anthropic is "racing straight to self-improving superintelligence and gambling with our lives." Fellow Anthropic researcher Evan Hubinger publicly backed him, writing: "Jacob is correct here... We really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade."
The story metastasized across the week. The Hill reported lawmakers sounding the alarm on AI dangers after the warnings, the BBC and CBS covered Hubinger's probability estimate, and NBC News reported Thursday on two AI researchers leaving Anthropic and Google over safety concerns, quoting one who said "there are no adults in the room." Reuters reported Thursday that lawmakers are now calling for new rules, and reports have also circulated that OpenAI has sought US rules governing powerful AI models.
From Warnings to a Bill File
What makes this week different from previous safety controversies is the speed of the policy response. Anthropic's own disclosures this week — a threat intelligence report detailing disrupted misuse campaigns, including possible attempts to use Claude toward developing biological weapons — gave the warnings an evidentiary backbone that previous debate lacked, and lawmakers appear to be moving while the moment holds.
The Road Ahead for AI Legislation
Prospects for the Frontier Act remain uncertain. Congress has struggled for years to pass comprehensive AI legislation, and any bill touching mandatory verification or a federal pause authority would face intense lobbying from the labs it would regulate, alongside jurisdictional fights between committees. On the sidelines, the states are not waiting: California Governor Gavin Newsom signed a suite of AI safety bills into law this week, per the Los Angeles Times, keeping up pressure for a federal framework that preempts a patchwork of state rules.
For now, the dynamic has inverted. For the first time in this Congress, the people building frontier models are publicly demanding restrictions on their own industry — and lawmakers are drafting them.
Whether that alignment produces law or merely headlines will depend on the coming months. But the week demonstrated something neither the labs nor their critics can ignore: AI safety has moved from the seminar room to the whip count, and the members counting votes say their constituents are finally the ones asking them to act.
AI Buzz Wire will continue tracking the Frontier Act as it moves through Congress.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
