Artificial intelligence has moved center stage in the upcoming midterm elections, propelled by a backlash against data centers and fears about AI safety. In response, Congress vows to act.
Democratic and Republican Members of Congress opened CEPA’s annual Tech & Security Conference with a rare sign of consensus: AI safety and innovation are not contradictory, insisted Representative Nathaniel Moran (R-TX-01). He compared the need for regulation to the rules that forced automobile companies to make cars safe.
“The brakes allow you to drive faster, not slower,” said Moran. “AI brakes should not slow innovation. They should enable faster movement with a safe backstop.”
This is the logic behind Moran’s proposed AI Kill Switch Act. Co-sponsored with Democratic Rep. Ted Lieu, it would require developers of the most powerful frontier AI models to be able to throttle, suspend, or completely shut models down if they threaten $100 million in economic damages or 10 deaths.
One fellow panelist, Rep. Lori Trahan (D-MA-03), introduced the bipartisan FRONTIER Act alongside Republican Rep. Jay Obernolte, both members of the House Committee on Energy and Commerce. The Act would impose a uniform, nationwide standard for catastrophic risk transparency, reporting, and auditing. The federal rules would override individual state laws, such as California’s independent safety audits, and eliminate a conflicting “50-state maze.”
“We need to have catastrophic risk transparency [and] incident reporting that is mandatory and not voluntary, and we need an independent verification model,” said Trahan.
The Frontier Act would establish an Undersecretary for AI Security within the Department of Commerce, and outlines a licensing framework for third-party independent evaluators. “Very large frontier developers” with more than $5 billion in annual revenue would be required to utilize a third-party evaluator to assess the risks of frontier models.
Trahan has also proposed the CLAIM Act to address AI liability. The act would make AI firms liable if their model breaks the law; courts would also assume that models are capable of intent. Similar standards already apply to drugs and cars.
The third panelist, Rep. Don Beyer (D-VA-08), distinguished himself by enrolling in a master’s degree program in machine learning at George Mason University to directly understand the technical foundations of the technology he regulates. He has floated a bill to establish minimum safety standards for advanced AI labs, including the establishment of a Federal Reserve-like structure to mandate testing of advanced models.
“There’s this whole spectrum of p-doom = 100 (the probability that AI will cause an existential catastrophe) and 0, and it’s probably somewhere in between,” Beyer said. “Everyone agrees there’s a real risk.”
All three pushed back against the Trump Administration’s hands-off tech approach. When administration advisers dismissed AI warnings as overly dramatic panic, Beyer countered that tech builders and Nobel Prize laureates have long highlighted AI’s risks to humanity.
All three encouraged engagement with the globe’s other AI power: China. They compared the need for a US-China deal with arms control treaties between the US and the Soviet Union in the 1970s and ‘80s. Panelists thought the Administration’s proposal for an AI incident “hotline” seemed insufficient as it is solely based on the “honor system.”
Congress is also concerned with how to regulate recursive self-improvement (RSI), when an AI model trains and improves itself without human prompting. A self-improving system could override a kill switch or cheat on AI safety testing. Congress will have to decide whether RSI is a red line not to be crossed.
Congress remains divided over how far regulation should go. While some are comfortable with taking down a model after a major incident, others prefer requiring a government agency to pre-approve all AI models.
But the three Members of Congress agreed that the US cannot win the technology race alone. European allies provide talent and infrastructure. Rather than competing, Europe and the US should cooperate.
Until recently, AI risks were theoretical, and the idea of snuffing out a promising technology with onerous regulation seemed like a mistake. Today, the “let it rip” approach looks outdated — both as a policy and, as American voters protest, a matter of politics.
William Echikson is a non-resident Senior Fellow at CEPA and editor of the Bandwidth Tech Blog.
Jack Galloway is a Program Assistant on CEPA’s Tech Policy Program.
Bandwidth is CEPA’s online journal dedicated to advancing transatlantic cooperation on tech policy. All opinions expressed on Bandwidth are those of the author alone and may not represent those of the institutions they represent or the Center for European Policy Analysis. CEPA maintains a strict intellectual independence policy across all its projects and publications.
Steel and Silicon: Allied Innovation for the Next Century
2026 CEPA Tech & Security Conference
