WASHINGTON — Sen. Rand Paul (R-Ky.) on Wednesday, Sept. 16, 2026, blocked a unanimous-consent request from Sen. John Kennedy (R-La.) for legislation requiring tech companies to build a “kill switch” into superintelligent AI systems, The Hill reported.

Kennedy argued Congress should balance innovation and public safety and warned that advanced models could, through recursive self-improvement, become an “independent species” that human operators cannot control. His bill would put companies — not the government — in charge of using an emergency shutoff if a serious danger arises. He asked colleagues for unanimous consent to pass it.

Paul objected. An “open-ended” mandate to implement kill switches, he said, could “ruin” AI’s rollout across the economy, and hasty congressional action risks killing innovation “and setting this technology back decades.” He asked whether search engines already using AI would have to be redesigned immediately, and whether doctors would need kill switches in drug-prescription analysis tools. Paul proposed amending the request to create a bipartisan committee to study superintelligent AI threats and options to stop an imminent danger.

Kennedy refused. A committee, he said, is “a way to kill this” and “the weenie way out,” adding he would rather listen to “OJ jokes for the rest of eternity” than set up another special panel. After Kennedy rejected the modification, Paul objected again — and the unanimous-consent push failed. The clash lands amid rising Capitol Hill alarm after recent OpenAI and Anthropic researcher warnings about existential risk. Mandating an open-ended “kill switch” before the technology is understood is exactly the kind of hasty rule that can freeze product design without proving it makes anyone safer.

Dated Wednesday, Sept. 16, 2026: Kennedy UC for company-run AI kill switch blocked by Paul; Paul sought study committee; Kennedy refused; no floor passage by UC.

Sources