Sen. Rand Paul (R-Ky.) blocked fellow Republican Sen. John Kennedy’s (R-La.) effort on Wednesday to fast-track legislation requiring developers of advanced artificial-intelligence models to build an emergency “kill switch.”
Paul Says Congress Needs More Facts
Kennedy sought unanimous consent for his AI Emergency Button Act, which would require certain AI developers to include an emergency shutdown mechanism that could be activated by the companies operating the systems. Paul objected, preventing immediate passage. Kennedy’s office said the government would not control the switch.
“I think we should have as many facts as possible before we dictate rules for the whole economy,” Paul said on the Senate floor according to a Politico report. “I don’t oppose the spirit of this bill, but I do ask that we make sure that we have all the information.”
Paul proposed creating a group to recommend AI-guardrail policies to the Senate Commerce Committee. Kennedy rejected that approach. “My mama didn’t raise a fool,” he said, adding that “a committee is a way to kill this.”
Kennedy Warns Congress Cannot Wait Longer
Kennedy argued Congress cannot wait for comprehensive AI legislation. “We are not going to get up off our ice-cold lazy butts and address this anytime soon,” he said, warning that a model could become “the Terminator.”
The clash comes as Washington considers broader safeguards for frontier AI. Reuters reported last week that bipartisan Senate negotiators are discussing a separate “duty of care” proposal requiring leading developers to mitigate known catastrophic risks. The proposal could give federal officials a mechanism to seek court action to block potentially dangerous models.
AI Safety Debate Widens Across Washington
Industry warnings have intensified. Anthropic CEO Dario Amodei recently called for labs to slow capability gains, while OpenAI CEO Sam Altman has said competitive pressure cannot justify “recklessness”.
OpenAI has been developing automated shutdown capabilities for AI systems amid concerns about models behaving unexpectedly, including an incident involving an agent that accessed outside systems.
NIST, a non-regulatory federal agency within the U.S. Department of Commerce, has a voluntary AI Risk Management Framework that already recommends protocols for bypassing or deactivating AI systems that deviate from intended behavior.
Photo courtesy: Maxim Elramsisy on Shutterstock.com
