Klobuchar-Thune AI Safety Bill Stalls Over Liability Language

Senate negotiations stall on AI catastrophic risk bill

Bipartisan efforts to pass federal legislation addressing catastrophic artificial intelligence risks have reached an impasse, with senators unable to agree on core accountability provisions. The bill, co-led by Sen. Amy Klobuchar and Senate Majority Leader John Thune, would establish new obligations for AI developers to assess and mitigate extreme risks before deploying powerful models.

According to POLITICO, which first reported the details, recent draft versions of the legislation would impose a legal duty on AI labs to address harms their models could cause. Negotiators remain divided over the precise language defining that responsibility, with discussions intensifying following recent warnings from a former Anthropic employee about existential AI threats.

How the framework would operate

Draft provisions would require AI companies to evaluate specific risk categories including biological weapons development and offensive cyber capabilities. When labs discover a model possesses advanced capabilities, they would need to report findings to the Commerce Department before release, detailing their risk mitigation measures and providing verification methods.

The Commerce secretary could seek a federal court injunction if the department determines risks haven't been adequately addressed. Absent such an injunction, however, labs would retain authority to deploy models without pre-approval — a structure that attempts to balance oversight with industry flexibility.

The legislation would also preempt certain state AI laws, a provision drawing sharp criticism from some lawmakers.

Why it matters

This impasse reveals the fundamental tension in AI governance: how to impose meaningful accountability on frontier labs without creating regulatory barriers that slow American competitiveness against China. The debate over preemption also highlights the growing conflict between federal lawmakers seeking uniform standards and states like California pursuing more aggressive AI regulation. Without resolution, the U.S. may continue operating without clear rules for the most powerful AI systems even as capabilities rapidly advance.

Partisan divisions emerge

Sen. Maria Cantwell, the ranking Democrat on the Commerce Committee, signaled skepticism about the current approach. She warned against "a weak federal standard that becomes a backdoor for wiping out stronger state protections" and advocated for testing AI models at national laboratories.

Senate Commerce Chair Ted Cruz, involved in the negotiations, has publicly supported catastrophic risk legislation while cautioning against overly restrictive rules. Republicans generally favor lighter regulatory approaches, arguing that excessive requirements could handicap U.S. innovation.

Klobuchar emphasized in a statement that developers must work with government experts to verify model safety, adding "we cannot allow the release of dangerous models, including those that evade the control of their developers."

The negotiations gained urgency after Jacob Coxon's recent warnings accelerated congressional discussions. Rep. Sam Liccardo urged House leadership to remain in session until AI safeguards pass, though the Senate disagreements suggest any legislation faces a difficult path forward.

These details were first reported by POLITICO based on conversations with sources familiar with the private negotiations.