President Donald Trump rejected calls for new restrictions on advanced artificial intelligence Monday, opening a sharper divide with technology executives who say government safeguards are falling behind rapidly improving systems. Trump argued that winning the global AI race matters more than slowing development and dismissed warnings that the technology could escape human control. His response came after leaders from Anthropic, OpenAI and other major laboratories urged stronger oversight, independent testing and more time for safety measures to catch up.
The dispute is no longer a narrow fight over corporate self-regulation. OpenAI has argued that voluntary commitments are insufficient for the most capable models and has called for national rules that scale with a system's abilities. The company has also disclosed that its newest systems require stronger cybersecurity safeguards and human supervision. Anthropic chief executive Dario Amodei has pressed for a slower pace of development, while Microsoft published principles aimed at preserving human control. Together, those moves have pulled questions about model testing, incident disclosure and legal responsibility into the center of national policy.
Congress faces a compressed calendar and deep disagreements over what a federal law should require. Lawmakers have discussed obligations for frontier-model developers, third-party evaluations, liability rules and limits on state regulation, but there is no settled package with a clear path through both chambers. The administration's earlier framework favored national standards while warning against rules that could weaken U.S. competition with China. The current clash makes compromise harder: industry leaders are asking for binding safeguards at the same moment the president is portraying that pressure as a threat to innovation.
The prediction market still treats enactment this year as unlikely. A Polymarket contract covering a federal AI-safety law with training, use or human-oversight requirements traded near fifteen and a half percent Monday afternoon. That was down three percentage points over twenty-four hours, although it remained six and a half points higher than a week earlier. About $107,000 has traded in the contract, with roughly $3,500 in the latest day. The price is an estimate of legislative odds, not a judgment about the underlying risks.
The next meaningful step is whether senators turn the latest warnings into bill text that can attract bipartisan support. Any viable proposal will have to define which systems qualify as frontier models, what testing must occur before release and who bears responsibility when an autonomous system causes harm. It will also have to bridge the administration's emphasis on speed with companies' demand for enforceable standards. Agencies would then need authority, technical staff and reporting systems capable of supervising models that change faster than ordinary rulemaking. Without that bridge, Congress may leave the year with hearings and frameworks but no binding law, even as developers say the window for governing the most capable systems is narrowing.
