| | By Benjamin Guggenheim | Not a subscriber? Sign up here to get this newsletter in your inbox. - Exclusive: Rep. Lori Trahan (D-Massachusetts) unveils key changes in her “Frontier Act 2.0,” which include a new federal floor on state AI laws and new language targeting automated AI research and development.
- Scoop: House Democratic leadership strategizes on how to move an AI package through as few committees as possible. Rep. Ted Lieu (D-California) looks to move jurisdiction for his kill switch bill.
- I interview Shaunna Thomas, the executive director of pro-worker super PAC Guardrails Alliance, on what Democrats should prioritize concerning AI in the new Congress.
This is Benjamin Guggenheim. Welcome to WP Intelligence’s AI & Tech Brief, where we examine the transformative technology of artificial intelligence at the intersection of innovation, policy and power. Get in touch at Benjamin.Guggenheim@washpost.com or on Signal at ben_guggenheim.88. Programming note: Stay tuned for a reported analysis on Anthropic’s public market debut by Lead AI & Tech Analyst Dan Gallagher in tomorrow’s edition. The stock opened trading this morning at $111.50, above its IPO price of $83 per share. That helped put the AI company’s market capitalization just under $2 trillion by the afternoon. The Lead Brief Rep. Lori Trahan (D-Massachusetts) is making key changes to her comprehensive AI safety legislation, which include establishing a federal floor for state AI laws and a new category of catastrophic risk targeting “recursive self-improvement” (RSI), which refers to automated AI research and development. “We can’t wait until January to start building strong, durable guardrails that can pass both chambers of Congress and be signed into law,” Trahan said in a statement on the newly released section-by-section of the Frontier Act 2.0, which I’ve obtained exclusively. “Building the coalition to get there starts now.” Other key changes to the Frontier Act include: - the creation of a new Office of AI Security within the Department of Commerce under a Senate-confirmed assistant secretary for AI security
- new civil penalties for “near misses” wherein an AI model incident would have likely led to loss of control or something catastrophic happening if not for luck or a “timely intervention”
- language that guarantees embedded access for independent verification organizations (IVOs)
- a broader definition of incident reporting that includes instances involving models deceiving their developers and critical infrastructure.
What it means: Overall, the changes make the Frontier Act stronger on AI safety — though it won’t be as welcome to Republicans who see preemption of AI state laws as key to any bipartisan deal. The original Frontier Act would have narrowly barred state laws covered by the new federal regime, including catastrophic risk disclosures, audits and incident reporting. But the new federal floor means that state laws that are more protective than the Frontier Act’s laws are preserved. The changes are a sure sign of how the political winds have shifted against the push for preemption spearheaded by House Majority Leader Steve Scalise (R-Louisiana) and venture capitalist David Sacks. It’s a reflection of both the new Democratic majorities in Congress that are set to take power, and the fact that states such as California are moving aggressively on new AI safety regimes. According to a spokesperson for the Massachusetts Democrat, the changes are a Trahan-led effort for now, but the lawmaker is continuing to talk about the changes with her partner on the original Frontier Act, Rep. Jay Obernolte (R-California), and other Republicans. But the field is getting crowded. Reps. George Whitesides (D-California), Nathaniel Moran (R-Texas) and Don Beyer (D-Virginia) announced the launch of the House AI Safety and Innovation Caucus on Thursday. And Beyer, Reps. Sara Jacobs (D-California) and Celeste Maloy (R-Utah) unveiled an outline of legislation that would have the government be primarily responsible for evaluating frontier AI developers. The Legislative Chessboard Meanwhile, Democratic leadership under Rep. Hakeem Jeffries (D-New York), the soon-to-be-speaker, is already strategizing on how to move a potential AI safety package through as few committees as possible, according to two people close to Democratic leadership, who spoke on the condition of anonymity to discuss the private conversations. Ideally, the vast majority of proposals for a federal AI regulation will be squeezed into the Energy and Commerce Committee, they tell me. Indeed, for this reason of committee jurisdiction, Rep. Ted Lieu (D-California), who was elected chair of the Democratic caucus on Wednesday, is looking to move AI model emergency shutdown authority, under his “kill switch” bill, from the Department of Homeland Security to the Commerce Department, according to the two people. That change will likely occur when Lieu reintroduces the bill in the next Congress. One thing that looks increasingly unlikely: The establishment of a new select committee on AI is looking to be an uphill battle. The incoming Democratic chairs of the committees are not thrilled about handing parts of their jurisdiction to a select committee, a senior House Democratic aide tells me. “Everyone with AI in their jurisdiction is doing everything to kill it,” said another House Democratic aide. “Sadly, I’m worried they’ll win out.” California moves aggressively: Trahan’s changes also come as California Gov. Gavin Newsom (D) unveiled a new road map on AI safety in Washington on Wednesday, which includes legislative recommendations for new authorities to cap recursive self-improvement. The road map calls for AI companies to undergo mandatory third-party evaluations. “The federal government’s reckless failure to create any form of meaningful AI oversight or accountability should alarm every American,” Newsom said in a statement. “We’re showing the country what responsible leadership looks like.” |