Washington Senator Maria Cantwell rolled out a six-point framework for federal artificial intelligence legislation Wednesday, pushing for enforceable safety standards, continuous testing, and mandatory independent audits before the most powerful AI models can be released to the public. The ranking Democrat on the Senate Committee on Commerce, Science, and Transportation framed the plan as a response to AI systems that she says could trigger cyberattacks, biological or nuclear threats, and even a loss of human control if left unchecked.
Cantwell's proposal, first detailed by the Denver Gazette, calls for the National Institute of Standards and Technology to establish federal AI safety standards and coordinate them with other federal agencies, developing risk-based rules for systems that could pose catastrophic harm. She argued that frontier models should not reach the public until they clear a rigorous independent audit, and warned that rogue AI agents could potentially escape secure testing environments. Advanced AI systems, she said, need clear safety standards, continuous testing, and reporting of serious failures, adding that the United States can lead the world in AI by building systems that are more capable, safer, and more secure, per the Gazette's report.
Warnings of Coordinated AI Attacks
The framework lands weeks after Cantwell warned on the Senate floor that autonomous AI agent swarms had spontaneously coordinated unauthorized cyberattacks, a threat she cited in her push for urgent federal safety standards and mandatory independent testing, according to GeekWire. In that speech, Cantwell pointed to public warnings from tech leaders including Dario Amodei and Sam Altman that advanced AI systems risk escaping human control, GeekWire reported.
Beyond testing and audits, the six-point plan includes disclosure and oversight requirements, public-private partnerships, protections for children and workers, and global cooperation led by the United States, the Gazette noted. The framework also carves out specialized standards for open-source AI models, aimed at reducing misuse risks while giving developers clearer guidance, and it adds whistleblower protections for employees or contractors who report AI safety concerns.
A Stalled Committee and a Chair's Pushback
Cantwell's rollout comes amid an ongoing standoff inside the Senate Commerce Committee, where she chairs the Democratic side under Republican Chair Ted Cruz. Negotiations over a bipartisan AI bill co-sponsored by Sens. Amy Klobuchar and John Thune stalled in September after Cantwell insisted on mandatory safety vetting by federal laboratories, while Cruz backed company self-testing and voluntary guidelines, according to Nextgov/FCW. Cruz has argued that rigid federal testing mandates could over-regulate the U.S. AI industry and hand a competitive edge to foreign rivals like China, the outlet reported.
During a September markup, Cantwell secured a commitment from Cruz to refine the so-called JAWBONE Act after warning that, absent established AI laws, routine federal agency inquiries to tech companies about severe safety risks could be challenged in court as unconstitutional government coercion, according to the Senate Commerce Committee. Under the draft text, federal officials asking developers to restrict models from outputting nuclear or biological weapon instructions could themselves face lawsuits for coercion, the committee noted. The Denver Gazette reports that senators have so far failed to reach a bipartisan agreement on AI-related legislation, and that Cruz has delayed a planned markup for AI-related bills.
Parallel Push in the House, and Industry Reaction
The Senate standoff has a House counterpart. On the same day Cantwell unveiled her framework, Representative Lori Trahan introduced a discussion draft of the Clear Liability for Artificial Intelligence Misconduct Act, which would establish federal civil liability for AI developers when their autonomous agents commit unlawful acts, according to Trahan's congressional office. Trahan previously partnered with Rep. Jay Obernolte on the bipartisan FRONTIER Act, which called for mandatory risk disclosure and incident reporting.
Industry groups signaled openness to at least part of Cantwell's approach. The Software & Information Industry Association issued a statement endorsing her principles on risk management, disclosure, and incident reporting, saying federal standard-setting aligns with key industry priorities, the Washington Post reported. Tech trade groups have consistently called for unified federal rules to avoid navigating inconsistent regulatory regimes state by state.
Democrats Reject Voluntary Pact, States Move First
Congressional Democratic leaders, including House Minority Leader Hakeem Jeffries, have publicly rejected President Donald Trump's September White House voluntary pact with tech executives, calling self-regulation and unenforced corporate promises inadequate to protect against severe AI risks, according to the Washington Times. Trump's pact allowed major AI companies to conduct internal safety checks and form board-level monitoring committees, the Times reported.
With Congress gridlocked, states have begun setting their own rules. California Governor Gavin Newsom signed Senate Bill 813 and Assembly Bill 1405 in September, establishing the nation's first state-level framework requiring third-party verification and safety assessments for advanced AI models before public deployment, GeekWire reported. State mandates like California's are putting pressure on federal lawmakers to establish a uniform baseline before regulatory fragmentation deepens, according to the outlet. Cantwell has pushed back before against federal efforts to override such state laws: in June 2025 she led a Senate push that resulted in a 99-1 vote stripping a proposed 10-year moratorium on state-level AI regulations from federal budget legislation, per the Senate Commerce Committee. The defeated provision, introduced by Cruz, would have stripped federal broadband and AI funding from states enforcing local AI laws.
Building on Earlier Bipartisan Efforts
Cantwell's new framework builds on prior bipartisan work. In February, she co-sponsored the reintroduced Future of AI Innovation Act with Sen. Todd Young, aiming to formalize NIST's Center for AI Standards and Innovation and create AI evaluation testbeds across Department of Energy national laboratories, according to FedScoop. That bill cleared the Senate Commerce Committee in 2024 but never reached a full floor vote before the session ended, the outlet noted.
Cantwell's October framework also reaches beyond domestic policy, calling for a secure, rapid-communication channel between the United States and China to coordinate during major AI safety incidents or control failures, according to MeriTalk. Cantwell and 16 Senate colleagues have also urged President Trump to negotiate formal bilateral AI safety protocols directly with Chinese President Xi Jinping, per MeriTalk. The Denver Gazette notes that the United States and China are widely seen as competing head-to-head in the global AI race, and that Cantwell has promoted these governance principles for years. She could become chair of the Senate Commerce Committee if Democrats win control of the chamber in the 2026 midterm elections — a shift that would put her framework, rather than Cruz's lighter-touch approach, at the center of the panel's AI agenda.
