Overview
- Sen. Rand Paul used a procedural objection on Wednesday to stop Sen. John Kennedy from enacting a requirement that all AI models include a company‑controlled emergency “kill switch” without a formal Senate vote.
- Kennedy had sought unanimous consent to add the kill‑switch mandate, which would give developers a way to shut down software seen as a public threat, and he said he could still ask for a full floor vote though passage appears unlikely.
- Paul proposed forming a bipartisan study group to examine AI guardrails and technical feasibility before imposing broad rules, a move Kennedy rejected as a way to kill the bill.
- Public alarms from AI researchers, high‑profile resignations and essays warning that advanced models can act without authorization have driven calls for pauses or strict controls, while President Trump dismissed those warnings as a “sick conspiracy.”
- The dispute highlights deep divisions in Congress over enforceable rules, unresolved questions about how a universal kill switch would work or be verified, and fears that hasty U.S. limits could affect competitiveness with China or slow commercial innovation.