Nancy Pelosi’s replacement is ready to regulate AI

Scott Wiener, the California state senator who authored the state's two flagship frontier AI bills, has won a tight congressional race for the San Francisco seat left open by former House Speaker Nancy Pelosi. He spoke with MIT Technology Review on Friday, a few hours after the result was declared, and said he was already focused on AI risk. He could now play a significant role in shaping federal AI policy.
Wiener pointed to a report published the previous day by the AI nonprofit METR, which he said had been widely read as evidence that Anthropic had achieved recursive self-improvement. He described that prospect as very concerning, and as a sign that the window for regulating AI is closing. He said he had hoped Congress would act decisively before the end of the year, and that this does not appear to be happening. Recursive self-improvement, the idea that a system meaningfully contributes to its own advancement so that progress potentially outruns control, came up repeatedly in the 20-minute conversation.
His legislative record gives him unusual standing on the issue. SB 1047, which he authored in 2024, was positioned to become the country's first real piece of frontier AI legislation before California governor Gavin Newsom vetoed it. Wiener then sponsored SB 53, a lighter-touch measure that Newsom signed. Both bills have inspired proposed federal legislation, which means his state-level drafting choices are already circulating in Washington in some form.
His campaign also drew scrutiny. Public First Action, a nonprofit that favors AI regulation and receives substantial funding from Anthropic, gave more than $3 million to his congressional campaign. Opponents used that to argue he is too close to the AI industry. Wiener rejected the charge, writing afterward that Public First backed him because he was the first lawmaker in the country to introduce AI regulation and the first to pass it and get it implemented, and he called the attacks shockingly cynical. For readers weighing that exchange, the relevant distinction is between a candidate funded by a pro-regulation group tied to a lab and a candidate captured by the industry he regulates; the source reports both the contribution and the denial without resolving them.
On policy, Wiener said he would be eager to serve on a congressional select committee on artificial intelligence, which Democrats have discussed since September. He wants a strong, empowered and nimble federal regulator and pointed to legislation from Representative Ro Khanna, a fellow California Democrat, that would create one. He also listed whistleblower protections, testing and auditing requirements, and a kill switch requirement as priorities, and said Congress should use its oversight powers to hold AI labs accountable, arguing the public needs both a say and visibility into what is happening.
He is firmly against broad federal preemption of state AI laws, which some proposed federal bills would require. Invoking Brandeis, he described states as laboratories that can stay nimble in ways Congress sometimes cannot. That position puts him at odds with industry arguments that a single national standard is simpler for compliance, and it suggests state rules like California's will remain a live patchwork rather than being folded into one federal regime.
Wiener frames the stakes in the vocabulary of the AI safety movement rather than in consumer-protection terms. He described the present as very dangerous with respect to loss of control over AI, and said recursive self-improvement has to be addressed because once systems are in that territory it is potentially a recipe for total loss of control. He said he wants Congress to act in a way that is truly meaningful and meets the moment. He counts Senator Bernie Sanders as a likely ally; Sanders has met with many influential figures in the AI safety movement and has spoken repeatedly about existential risk.
The wider political terrain looks less aligned. Voters approved 40 of 41 anti-data-center ballot measures on Tuesday, which could be read as public appetite for restraining AI, but Wiener sees that movement as driven mainly by local concerns: impacts near people's homes, higher electric bills and water use. His reading is that the revolt is about siting and utilities rather than about catastrophic risk, which matters for anyone assuming hostility to data centers translates into support for safety legislation.
He describes himself as only cautiously optimistic about Congress's ability to rein in AI, citing its poor track record on regulating the tech industry. He also raised the obstacle of President Donald Trump, who in September called concerns about catastrophic AI risk a hoax and has shown a stark unwillingness to collaborate with China, which Wiener considers essential to preventing the worst outcomes. He suggested Trump's unpredictability offers a fraught kind of hope, noting that what has been seen so far are problematic things but not complete disasters, and that he hopes no disaster is needed to prompt action, but if one is, that it be as small as possible.
This suggests the practical question for AI companies is not whether a sympathetic legislator has arrived but whether a divided Congress can pass anything with teeth. Wiener's own bills show the likely path: an ambitious version that draws industry opposition and gets vetoed, followed by a narrower version that becomes law. Federal preemption remains the key fault line, because a national standard could either raise floors across states or wipe out the stricter rules Wiener spent years building. The kill switch, auditing and whistleblower proposals he favors would, if enacted, translate into concrete engineering and documentation work for labs.
Why it matters: Developers and product teams selling or deploying frontier models in the US face a legislator with a track record of turning safety rhetoric into statutory obligations, and his arrival in Congress keeps state-level rules like California's relevant rather than superseded. Inference: expect continued uncertainty over whether compliance should track state law or await federal action, and expect the auditing and testing requirements he endorses to be the provisions most likely to shape real engineering practice if any bill advances.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
As a California state senator, Scott Wiener authored legislation aimed at curbing the risks of advanced AI. Now, after winning a tight congressional race for former Speaker of the House Nancy Pelosi’s open San Francisco seat, he could play a key role in setting federal AI policy. I spoke with Wiener on Friday, only a…