WASHINGTON — A bipartisan Senate effort to regulate the most dangerous uses of artificial intelligence remains stalled over how much responsibility to place on AI companies and how much authority to give the federal government. The legislation, led by Sen. Amy Klobuchar, D-Minn., and Senate Majority Leader John Thune, R-S.D., is still being negotiated and has not been finalized. People familiar with the discussions said lawmakers remain divided over provisions that would impose a legal duty on AI labs to reduce harm caused by their models. Draft language circulated in recent days would require companies to assess risks including the creation of biological weapons and the use of AI to conduct offensive cyber operations. If a lab discovers that a model has advanced capabilities, it would have to notify the Commerce Department before releasing the system and provide details about safeguards designed to control the danger. The Commerce secretary would be allowed to seek an injunction in federal court if the department determined that a risk had not been adequately addressed. The proposal would not create a blanket federal pre-approval system: absent a court order, companies could decide whether to release a model. That approach reflects an attempt to balance public safety with concerns that Washington could smother a fast-moving American industry. Republicans, including Senate Commerce Committee Chairman Ted Cruz, R-Texas, have generally favored a lighter regulatory touch, warning that excessive federal rules could slow U.S. innovation and give China an advantage. “It is possible to move bipartisan legislation,” Cruz said on ABC’s “The View.” “But it ain’t easy.” The negotiations accelerated after warnings from former Anthropic employee Jacob Coxon that artificial intelligence could pose an existential threat within the decade. Rep. Sam Liccardo, D-Calif., separately urged Speaker Mike Johnson to keep the House in session until Congress approves AI safeguards. Another major fault line is federal preemption. The drafts would override some state AI laws, a provision Democrats on the Commerce Committee have resisted. Sen. Maria Cantwell, D-Wash., the committee’s top Democrat, said a federal standard should not become “a backdoor for wiping out stronger state protections.” Cantwell also called for AI models to be tested by scientists and experts at national laboratories before they are deployed. Her position would add a more demanding layer of government review than some Republicans and industry advocates appear willing to accept. Klobuchar said she remains committed to reaching an agreement and supports requiring developers to work with government experts to test and verify advanced models. “We cannot allow the release of dangerous models, including those that evade the control of their developers,” she said. The unresolved questions are consequential for businesses, consumers and state governments. A federal system with clear, narrow rules could provide companies with a uniform national standard instead of a patchwork of state mandates. But critics of preemption say weaker federal language could strip states of protections they adopted in response to risks Washington has yet to address. For now, the bill’s fate depends on whether senators can settle those competing demands without turning the proposal into either an industry-killing regulatory regime or another largely toothless promise from Congress.