A bipartisan House duo is racing against the clock to build consensus around a framework for regulating artificial intelligence — and prove Congress is up to the task of setting guardrails around the rapidly-evolving technology.

Reps. Jay Obernolte (R-Calif.) and Lori Trahan (D-Mass.) formally introduced legislation last Thursday with four cosponsors — two from each party — that would give the government more oversight over cutting-edge AI models as new and potentially dangerous technological advances continue to go largely unchecked.

It’s a considerably reworked version of a discussion draft the two lawmakers unveiled weeks earlier. It was highly anticipated as the most serious cross-party attempt at comprehensive AI regulatory legislation to date — and almost immediately panned by Republicans and Democrats alike.

But Trahan and Obernolte both think they’ve now found the sweet spot and are hopeful the Energy and Commerce Committee schedules a hearing or markup on the measure when the House reconvenes in September after the summer recess.

“That is the goal,” Trahan said in an interview Friday. “I think that there is real urgency. There’s earth-shattering news practically every week, and that’s only going to get more frequent and more serious as these models get more powerful.”

Obernolte said in an interview earlier in the week he “absolutely” thought the bill could move out of committee in the fall. “We’re certainly going to try and get it done this year.”

Yet when asked last week about whether he would bring up that bill — or any comprehensive AI legislation — before the midterms, Energy and Commerce Chair Brett Guthrie was noncommittal: “I’m not sure if we’ll get there before the election,” the Kentucky Republican said.

Congress has ultimately done little to legislate on AI under the Republican governing trifecta, as members remain divided on how to tackle the thorny policy area — particularly the question of how involved the federal government should be in regulating private sector innovation. The Obernolte-Trahan bill is the latest test for whether lawmakers can reach consensus on the issue before the November elections.

In the 119th Congress so far, lawmakers have managed to send to the president’s desk one bill that criminalizes the sharing of AI-generated deepfakes and “revenge porn.” The Energy and Commerce Committee has passed bipartisan bills on data centers and kids online safety but none have advanced onto in the Senate.

Across the Capitol, the Senate Commerce Committee was just forced to delay a highly-anticipated markup of AI legislation at least a week as Chair Ted Cruz (R-Texas) struggles to put together an agenda of bills that stand a chance of passing on the chamber floor.

In the House, Obernolte and Trahan are trying to set their proposal up for success despite the odds. In interviews, they both described how they worked hard to respond to the initially negative feedback they received from colleagues on their discussion draft.

They have seriously scaled down provisions in their latest version relating to preemption of state AI laws, seeking to strike a balance for lawmakers who are split between those who want the ability to replace local regulations with a federal framework governing AI, and those who want to leave regulatory action to the states.

“We all heard the criticism … and we narrowed the preemption [provisions] substantially,” Trahan said. “It now only touches three areas the federal bill actually covers: catastrophic risk transparency, independent verification and incident reporting. That’s it.”

“So, child safety laws are explicitly protected,” she continued. “So is the consumer protection, privacy and state procurement. All of that stays with the states.”

Obernolte acknowledged the topic of preemption makes people “emotional” but the new legislative language more precisely describes exactly “what we meant on what we were preempting … and I think that that’s giving everyone more comfort.”

Their bill would propose robust emergency shutdown authorities for AI models determined to be major national security risks, replacing the more haphazard nature of the Trump administration’s response in recent months to concerning technological developments.

It also would require large developers to publish safety reports, report safety incidents and participate in independent safety audits. Those auditors would have to then report any models to the government that pose “imminent catastrophic risk” and the Secretary of Commerce could choose to restrict their deployment.

Last week’s first documented case of an autonomous AI breach helped bolster their case that this kind of legislation is necessary, both lawmakers said, with Trahan pointing out, “we only know about these incidents because OpenAI chose to tell us — there’s no law that requires it.”

Trahan, a member of House Democratic leadership, has been working outside the structure of the three-member AI commission overseen by Minority Leader Hakeem Jeffries but was optimistic when asked if the proposal might eventually get their backing.

“The feedback has been overwhelmingly positive,” she said, adding she plans to spend recess talking with party leaders and the top Democrats on key committees “whether there’s a path to make this bill even better and to move it before we break for the midterms.”

Obernolte said he’s talking to GOP leaders “constantly” to “make sure that everyone knew what was going on and what we were doing.”

But he also added that he hasn’t yet asked Republican leaders for their endorsement, arguing it would be “inappropriate” for them to give the bill their blessing “at this point.”

In the hours before Obernolte and Trahan released their formal bill text, Majority Leader Steve Scalise said in an interview that while he had not spoken to Obernolte about what the two lawmakers intended to introduce, Republican leaders were interested in advancing a regulatory framework that they deemed not overly burdensome.

“We’re trying to get AI preemption … but we also want to make sure that we don’t impede innovation with heavy regulation from the government,” said Scalise.

A Scalise spokesperson did not respond to a request for comment about what the Louisiana Republican thought of the latest proposal.

House Speaker Mike Johnson has also not taken a position on the bill. His last comments about the Obernolte-Trahan effort were in June, when he said he was planning to discuss the then-forthcoming regulatory framework with OpenAI CEO Sam Altman during the tech giant’s visit to the Hill.

Obernolte is confident his bill with Trahan represents the emerging consensus view about where lawmakers need to go next on AI policy, even within the Republican Party that has long championed limited government regulation.

“I think that everyone would agree we need food safety inspectors, right? So sometimes the two parties disagree on just how onerous to make those regulations. … I don’t think anyone’s saying that we shouldn’t have any regulation of AI,” he said.

He also suggested perspectives may be shifting.

“I think everyone’s thoughts on this topic have evolved as the technology itself has evolved, and because AI is such a fast-moving technology, I think that Congress is going to have to stay very engaged on this,” Obernolte said. “So, even if we succeed in passing [the bill] this year and getting it signed into law, immediately in the next Congress, we’re going to have to dive into the things that have changed.”

In a sign Obernolte may still have people to convince inside his party, though, Scalise last week rejected the idea that Republican views on the subject have evolved.

“No,” Scalise said. “I mean — our Republicans, by and large, are against heavy regulation that impedes innovation in any industry.”

Source link