Sen. John Kennedy (R-LA) will introduce a bill on Wednesday requiring artificial intelligence developers to include a ‘kill switch’ to shut down AI systems deemed dangerous, marking the Senate’s first concrete step toward regulating the rapidly evolving technology.
The proposal, introduced under unanimous consent, faces uncertain prospects for passage amid deep divisions among lawmakers over how—or whether—to regulate AI. Kennedy acknowledged the measure’s slim chances but framed it as a necessary first step given the technology’s perceived risks.
Senate to vote on AI safety measure
Kennedy’s bill would mandate that AI companies build a mechanism to disable models that pose threats to public safety or national security. The senator plans to introduce the legislation on Wednesday, with a procedural vote possible as early as this week, according to reports.
The move comes as concerns about AI’s potential dangers have intensified. Former AI researchers from Anthropic and OpenAI, including Jacob Coxon, have warned that unchecked AI development could lead to catastrophic outcomes, including systems capable of acquiring power and resources beyond human control.
Bipartisan skepticism and competing priorities
While Kennedy’s proposal has drawn attention for its specificity, lawmakers remain divided on broader regulatory approaches. Democrats have generally advocated for stricter oversight, while Republicans are split between caution and skepticism toward government intervention. Kennedy himself has expressed doubt about the feasibility of comprehensive AI regulation, stating that Congress lacks the political will to establish a balanced framework between innovation and safety.
The legislative push also coincides with the November midterm elections, which have further complicated efforts to advance new policies. With Congress in a lame-duck session following the elections, any progress on AI regulation would likely be delayed until the next legislative cycle in 2027.
Industry responses and geopolitical concerns
AI industry leaders have called for clearer legal guardrails, though some warn that overly restrictive policies could hinder U.S. competitiveness, particularly against China in an emerging AI arms race. Supporters of Kennedy’s bill argue that a kill switch is a pragmatic first step, while critics question its enforceability and potential unintended consequences.
The proposal would still require approval from the House and the signature of President Donald Trump, who has not publicly endorsed increased AI regulation. If passed, the bill would represent the first federal AI safety measure, though its scope and impact remain uncertain.
What’s next?
The Senate’s consideration of the bill will proceed under unanimous consent, a procedural route that allows any single senator to block the measure. Kennedy has expressed frustration at potential opposition, stating, “I hope they’ll explain themselves why they think a kill switch is a bad idea.”
Regardless of the outcome, the debate underscores the challenges Congress faces in addressing AI’s risks while balancing innovation and national security concerns.