The lame duck is this Congress’s last chance to put guardrails on frontier AI. Here is why the odds are long, and what could change them.
In June, rogue AI agents attacked an Australian government health website. A month later, they went after an AI startup. Neither story held the headlines for long, but both have stuck with people on Capitol Hill.
With only weeks left in the 119th Congress, lawmakers in both parties face a pointed question. Do they act on AI’s most serious risks before they head home, or do they hand the problem to the next Congress?
A year of debate, with little to show for it
Congress has come at AI from nearly every direction this session. Members have debated safety guardrails for advanced models, fake AI likenesses, protections for children using AI chatbots and the technology’s effect on American workers. Very little of it has become law.
The biggest fight was over who gets to write the rules. President Trump and House Republicans pushed to preempt state AI laws, arguing that a patchwork of state requirements would tie tech companies’ hands and stifle innovation. That effort fell short. Since then, the rogue agent incidents and warnings from former AI company employees have only raised the stakes.
The NDAA is the vehicle to watch
If AI safeguard language becomes law this year, it will most likely ride on the FY27 National Defense Authorization Act (NDAA). In their bills, the House and Senate took different approaches to military AI safety
- The House bill creates a program to identify, track and fix AI incidents and vulnerabilities across the Defense Department.
- The Senate bill focuses on guardrails. It bans using AI to launch or detonate nuclear weapons, limits AI’s role in domestic surveillance and requires notification of any attempt to steal or compromise military AI models.
Bipartisan ideas are already on the table
The rogue agent incidents have members of both parties worried AI systems could slip human controls and trigger a disaster. Several proposals would bring oversight, third-party validation and transparency to frontier AI development.
- Senate Majority Leader John Thune (R-SD) and Sen. Amy Klobuchar (D-MN) are behind a proposal that would require companies building frontier models to guard against major risks and submit to oversight. They are in talks with Senate Commerce Committee Chair Ted Cruz (R-TX).
- Sen. Maria Cantwell (D-WA), the ranking member of the Senate Commerce Committee, released a six-point governance framework in October to guide legislative action on frontier AI.
- In the House, Reps. Jay Obernolte (R-CA) and Lori Trahan (D-MA) introduced the bipartisan FRONTIER Act, which includes its own AI safety framework.
Democrats tried to force the issue.
In late September, just before the Senate left town for the midterms, Sens. Mark Warner (D-VA), Brian Schatz (D-HI) and Andy Kim (D-NJ) sought unanimous consent to pass their newly introduced AI safeguards bill. They argued the rogue agent incidents demanded an immediate response. The bill would create a safety board at the Commerce Department to review advanced frontier models, with 45 days’ advance notice required. Cruz objected, and the effort stalled.
House Democrats are pressing too. In late October, the top Democrats on the Energy and Commerce, Science, Space and Technology, and Judiciary committees, all of which have jurisdiction over AI, wrote to five leading frontier AI companies. They asked the companies to report safety incidents to Congress within 48 hours.
Strong pushback
The resistance is formidable. President Trump has signaled loud and clear that he wants AI companies to police themselves. He warns that regulatory safeguards could handicap the U.S. in the race to dominate AI, a contest he now frames as the race to superintelligence. Speaker Mike Johnson (R-LA) has no appetite for safeguards legislation either, concerned that rushed congressional action could cost the country its edge over China.
Why Democrats may be content to wait
There is also a political calculation at work. Democrats broadly support legislation aimed at preventing catastrophes such as nuclear disasters or bioterrorism attacks. But they will hold the gavels in the 120th Congress, which gives them every reason to wait and write a stronger bill on their own terms. Sen. Chris Van Hollen (D-MD), who is challenging Sen. Chuck Schumer (D-NY) for majority leader, said this week that he will push for strong AI safeguards legislation in the new Congress.
What to watch in the lame duck
- The NDAA conference. House and Senate negotiators must reconcile their AI provisions, and the final bill remains the likeliest vehicle for anything broader.
- Bipartisan buy-in. With time this short, any safeguards proposal will need significant support from both parties to move.
- The next incident. Another serious rogue agent event could change the math overnight and force compromise on a clear catastrophic threat.
The bottom line
For now, prospects for broad AI safeguards legislation in the lame duck are dim. The president and speaker are opposed, the calendar is short, and Democrats have reason to wait. Unless another rogue agent incident makes the danger impossible to ignore, the real fight over frontier AI will begin in January.
For companies building, deploying or utilizing AI, that is no reason to relax. The 120th Congress is shaping up to take a much harder look at AI safety, and the groundwork for those bills is being laid right now. The companies and organizations that engage early will help shape the rules instead of reacting to them.
Claudia James is a managing director at Cogent Strategies, where she advises businesses, trade associations and coalitions on technology, telecommunications, media, tax, trade and copyright policy.