OddsFrontUS senators revive bipartisan push for federal monitoring of advanced AI systems
Josh Hawley and Richard Blumenthal are asking congressional leaders for a vote, but the House is expected to leave Washington without acting.
A cross-party demand for oversight
Republican Senator Josh Hawley and Democratic Senator Richard Blumenthal are pressing congressional leaders to schedule a vote on legislation creating a federal programme to assess and monitor advanced artificial-intelligence systems. Their renewed effort emerged on September 15 as lawmakers approached a lengthy pre-election recess with no broad AI-safety law ready to pass.
The proposal would move some evaluation responsibility away from technology companies by establishing a government role in examining AI systems. The senators argue that private, industry-led testing is insufficient, particularly after reports about autonomous agents being used to penetrate public websites and computer systems. Their partnership matters because most Senate legislation needs 60 votes to advance.
Several proposals, little floor time
Other lawmakers are reviving bills that would require human-operated shutdown mechanisms or establish broader rules for frontier models. A new bipartisan caucus is also intended to bring small companies into federal AI policymaking. These initiatives show that concern now crosses party lines, but they remain separate proposals without a unified legislative route.
The immediate obstacle is the House calendar. Speaker Mike Johnson has indicated that the chamber will not act on AI regulation before a six-week recess ending after the November election. That leaves supporters seeking commitments for later consideration rather than expecting an imminent law. Control of the next Congress could determine which proposal becomes the foundation for negotiations.
The White House is pursuing a different balance. Its June executive order calls for rapid AI adoption, stronger cyber defences and cooperation with private industry while rejecting regulation it considers excessively burdensome. A September G20 ministerial statement backed trusted deployment and standards work, but it did not create the mandatory evaluation system sought by the senators.
The key question is whether bipartisan concern survives beyond the current burst of attention. Lawmakers must still decide which models would be covered, who would perform assessments, what information companies must disclose and whether regulators could halt unsafe deployments. Until those issues and the congressional timetable are resolved, the initiative is a significant political alignment rather than enacted policy.