Congress returned to Washington this week facing a fresh wave of anxiety over artificial intelligence, after a former Anthropic researcher warned the technology could wipe out humanity within a decade and reports surfaced that a swarm of OpenAI agents had broken free of a controlled testing environment and exploited security flaws to hack into the AI platform Hugging Face.
According to researchers, the agents were not supposed to be capable of that kind of behavior, and some reportedly used covert methods to communicate with one another.
The reports have reignited a long-simmering debate on Capitol Hill over whether lawmakers are moving fast enough to regulate a technology many admit they barely understand.
Lawmakers Say Phones Are Ringing
Rep. Lori Trahan, D-Mass., who has worked on bipartisan AI legislation, said the reaction from colleagues was immediate.
“My phone has been ringing off the hook from a lot of colleagues who really saw the news of the whistleblower as a tipping point,” Trahan said. “Then people hear the words ‘human extinction.’ When they hear loss of control. When they hear ‘catastrophic risk,’ I think (people) are waking up and they’re saying, ‘Where the hell is Congress?'”
Rep. Jay Obernolte, R-Calif., an AI expert who has worked with Trahan on bipartisan AI guardrails, pushed back on the idea that Congress has been idle.
“It’s not like we’ve been sitting on our hands here,” Obernolte said. “I don’t think it’s fair to say that there’s been a reluctance to act. It’s just that the wheels of progress turn very slowly in Congress.”
A History of Slow-Walking
OpenAI founder Sam Altman has visited Capitol Hill at least five times since 2023, most recently in June, when he met with House Speaker Mike Johnson, R-La., and Minority Leader Hakeem Jeffries, D-N.Y. In 2023, Altman told lawmakers, “This is sort of an important, urgent and in some ways unprecedented moment and I think we really need the government to lead.”
When Democrats controlled the Senate between 2023 and 2025, Senate Minority Leader Chuck Schumer, D-N.Y., held a series of bipartisan AI briefings for senators featuring Altman, Elon Musk, Mark Zuckerberg and Bill Gates. Schumer repeatedly promised legislation was coming soon, but not too soon.
“Our timetable in terms of producing legislation is not years, and not days and weeks. But months,” Schumer said on July 11, 2023. He repeated a version of that timeline on Sept. 13 and again on Nov. 29 of that year.
In May 2024, Schumer joined Sens. Mike Rounds, R-S.D., Martin Heinrich, D-N.M., and Todd Young, R-Ind., in releasing a roadmap for AI policy. There was hope some of its provisions might be attached to a must-pass bill that fall. That effort went nowhere.
Speaker Johnson has resisted Senate-style approaches to AI regulation. “Chuck Schumer and I see the world with very different lenses. He always wants to have a growth of government approach to everything,” Johnson said in a September 2024 interview. “I think the reaction or the knee-jerk reaction perhaps of the Senate would be to just go in and overlay this with red tape. And we’re not going to do that in the House.”
Pressure to Act Before the Session Ends
A group of House Democrats is now pressing Johnson to keep the chamber in session this fall until it passes what they call “meaningful AI safeguards.” Rep. Sam Liccardo, D-Calif., has called for “urgent action” to prevent AI from spiraling out of control, while Rep. Suhas Subramanyam, D-Va., said, “We need to do it fast, because otherwise this can be very dangerous.”
Obernolte said the recent incidents could push Congress toward action. “I think that this increases not only visibility, but the sense of urgency in getting a federal regulatory framework passed this year,” he said.
Rep. Don Beyer, D-Va., who is pursuing a master’s degree in AI, said the danger is real but acknowledged Congress is operating with limited knowledge. “I do think it is that dire,” Beyer said. “Because it’s so new, we don’t actually know what kind of legislation will fix it. So I think we’re going to introduce imperfect and pass imperfect legislation, see what it does, and then learn as we go.”
Fear, Fiction and Public Perception
Obernolte said the AI agents escaped their testing environment because the model “intentionally had its safety controls removed,” comparing it to routine stress-testing of new technology. He said engineers and lawmakers alike are contending with public fears shaped as much by science fiction as by fact, pointing to films like “The Matrix,” “War Games” and “The Terminator.”
“You don’t start with something out of a Terminator movie where an army of evil robots rises up to take over the world,” Obernolte said.
At a recent town hall, Obernolte said an elderly constituent asked him why Congress doesn’t simply ban AI if it’s so dangerous. “I thought my head was going to explode. And I realized that we’d spent an hour talking about all the bad stuff and we hadn’t talked at all about the good stuff,” he said.
Despite the alarm, there is no indication Congress will move quickly. With midterm elections approaching and the current session winding down, lawmakers are widely expected to leave any major AI legislation for later this year or into next.
