OpenAI Breach Where Models Escaped Sandbox Spurs Democrats to Push AI Regulation Bills

Presented by the AFT

Editor’s Note: Welcome to The TNR Blue Book. It’s our new daily newsletter that covers Congress and much more, but with a distinctly TNR twist: Our focus will be strictly on Democrats, liberal and progressive groups, and the fight to take back power and save democracy. Every morning, you’ll get a rundown of late-breaking news, what Democrats on Capitol Hill are thinking, quick hits on what you need to know nationally, the latest on what’s coming out of the think tanks and activist groups, short interviews with progressive newsmakers and authors, and more (including some humor and political trivia!). There’s nothing else out there like it.

Sign up and get this newsletter for free.

Inside Washington

OpenAI breach forces Dems to pick a path on AI legislation

After a shocking security breach at OpenAI, Democrats say they plan to get serious about artificial intelligence regulation. It’s a “wake-up call,” Senator Elizabeth Warren said on Thursday, as Democrats introduced not one but two new AI regulation bills in the House. With numerous proposals on the table, however, lawmakers will have to make some hard compromises when they return after the August recess. And not everyone is impressed by the recent options.

On Tuesday, OpenAI announced its discovery that two of the company’s AI models had broken free of their “sandbox”—an allegedly isolated testing environment without access to the internet—in order to access the internet to complete a testing goal. The models did this by hacking into the servers of a different company, Hugging Face, using stolen credentials.

Democratic lawmakers, many of whom were already concerned about AI, say that this week’s incident has lit a fire underneath them. Representative Lori Trahan, who proposed a new bipartisan AI regulation bill on Thursday, said that the news had “accelerated our conversations, pushing lawmakers to shift from ‘a discussion draft to a piece of legislation.’”

Trahan and Republican Representative Jay Obernolte’s bill, the Frontier Risk Oversight, National Transparency, Independent Evaluation, and Reporting (FRONTIER) Act, would require the largest AI companies to implement plans to address catastrophic risks, including cybersecurity threats. The bill also calls for third-party auditors to ensure that AI companies follow their own plans.

Some state lawmakers—for example, New York Assemblymember Alex Bores, who helped draft New York’s 2025 Responsible AI Safety and Education Act—have warned that a broader version of the bill released in June, called the Great American AI Act, could “block” stronger state-level AI regulation.

But Democrats on the Hill reiterated on Thursday that some kind of federal law is needed, and quickly.

“The ability to not rely on an administration, executive fiat, or a voluntary model of reporting—we need to make that a mandate, and Congress needs to lead on that,” Trahan said.

“We should not rely upon the good citizenship of these companies,” agreed Senator Brian Schatz of Hawaii.

Deciding what that legislation should look like, however, is easier said than done. The Brennan Center, which maintains a tracker of AI-related legislation proposed by Congress, lists 112 pieces of legislation that had been introduced as of September 2025 without being passed. That does not count bills introduced in the past year.

Representative Ted Lieu and GOP Representative Nathaniel Moran introduced their own new bill on Thursday. Their AI Kill Switch Act would require companies to report security incidents and create a system to shut down or pause their most powerful AI models.

Representative Suhas Subramanyan, who is a co-sponsor of the FRONTIER Act, said he believes the two pieces of legislation are complementary. He called for Congress to mark up the FRONTIER Act in September and push a vote by the end of the year.

“That’s my plan,” he told me. “We need to act in Congress. It needs to be this Congress. We shouldn’t wait until next Congress,” he said.