Key Highlights
- A researcher's resignation from Anthropic, warning that AI could cause human extinction, has intensified political attention on AI safety in Washington.
- Progressive and conservative lawmakers found rare common ground at a Washington gathering calling for urgent AI legislation.
- Multiple competing federal AI safety bills are circulating in Congress, differing on scope and enforcement mechanisms.
- Polling shows a large majority of Americans support AI regulation even if it slows innovation, though political consensus on specifics remains elusive.
A Resignation That Shifted the Debate
A wave of political attention toward artificial intelligence safety has followed the public resignation of a researcher at Anthropic, who said he and colleagues believed AI could pose an extinction-level risk to humanity within years. The researcher, who had worked on pretraining research at both Anthropic and OpenAI, said in his resignation message that neither company was acting responsibly and that the industry was "racing straight to self-improving superintelligence" without adequate safeguards.
The resignation, which spread widely after being shared on social media, has been credited with catalyzing a broader political reaction. Days later, a gathering in Washington, DC, called the Pro-Human Assembly brought together an unusual coalition of figures from opposite ends of the political spectrum, warning of the unchecked power of technology companies and calling for urgent legislation to slow AI development.
Political Reaction Splits Along Familiar Lines
Despite growing bipartisan attention to the issue, deep divisions remain over how to respond. Democratic leaders in the House and Senate called for urgent action, while the Republican speaker of the House warned against hasty regulation, arguing that overly aggressive congressional action risked ceding ground to China's AI development. He said AI companies should largely be left to "self-police."
Polling suggests the public leans toward regulation. Roughly eight in ten Americans say they support AI oversight even if it slows the pace of innovation, according to recent survey data, though there is little consensus on what specific rules should look like.
A Patchwork of State and Federal Proposals
Most regulatory activity to date has occurred at the state level. California enacted a law last year requiring AI companies to report safety incidents within 15 days, backed by financial penalties, and New York followed with its own disclosure requirements. Advocates say both laws would have been stronger were it not for opposition from industry-backed political groups.
At the federal level, lawmakers have introduced a range of competing proposals. One bill would require third-party evaluation of AI models and allow regulators to suspend systems posing imminent catastrophic risk, defined in the bill as causing more than 50 deaths or at least $1 billion in damages. A separate bipartisan Senate proposal would allow the government to seek a court order blocking the release of a model, while another, more stringent bill would require developers to submit models for government testing before release.
Industry's Mixed Signals
Several AI companies have voiced conceptual support for some form of oversight. One company has pointed to aviation and food safety regulators as potential models for AI governance, while another has floated an industry standards body backed by government oversight. However, no major AI company has fully endorsed any of the sweeping state or federal proposals currently under consideration, aside from limited support for California's disclosure law.
Even within the administration, officials have offered differing signals. While the president dismissed AI safety concerns as a "hoax" in Social Media posts, the vice president suggested days later that some firms building particularly risky systems should voluntarily halt development, a comment that stood somewhat apart from the administration's broader anti-regulatory posture.
Outlook
Whether federal AI legislation advances before the midterm elections will likely depend on whether lawmakers can bridge disagreements over the scope of any bill, including whether it should address only catastrophic risks or also cover issues such as Liability and children's use of AI chatbots. Advocates say political positioning on AI regulation is likely to harden as the 2028 presidential campaign approaches, though any binding rules in the near term would require agreement between Congress and the current administration.






Please wait processing your request...