The New Representative in Pelosi's Former Seat Wants to Give AI Regulation Real Teeth

Scott Wiener has won the vacant congressional seat previously held by Nancy Pelosi in San Francisco. Just hours after the election result was announced, he sat down for an interview with MIT Technology Review, with the conversation already centered on AI risks.

He previously spearheaded AI legislation in the California State Senate. SB 1047 introduced in 2024, which would have become the United States' first true frontier AI regulation bill, was vetoed by Governor Gavin Newsom. He later pushed for the more moderate SB 53, which Newsom signed into law. Both bills have since inspired federal-level legislative proposals.

The day before the interview, non-profit organization METR released a report widely interpreted as proof that Anthropic had achieved recursive self-improvement — a concept describing AI's ability to materially advance its own capabilities, which could cause technological progress to spiral outside human control. Wiener called this development "extremely alarming". He originally hoped Congress would take decisive action before the end of this year, "but that does not look like it will happen". Over the 20-minute conversation, he mentioned the term "recursive self-improvement" six times.

His assessment is straightforward: the pace of AI progress is so fast that we need to hit the brakes, otherwise people will sleepwalk into a catastrophe.

On policy, he supports a strong, well-mandated, agile federal AI regulatory agency. Ro Khanna, the Democratic U.S. Representative from California, has already introduced legislation to establish such a body. Wiener also emphasized whistleblower protections, mandatory testing and audit requirements, universal emergency kill switches for high-risk AI, and congressional oversight to hold AI labs accountable. As he put it, this concerns the future of humanity, so the public deserves to be informed and to have a say in the process.

He opposes broad federal preemption of state-level AI laws. As a long-serving state legislator, he cited Louis Brandeis' famous framing of U.S. states as "laboratories of democracy", noting that state governments are often more agile than Congress when rolling out new regulations.

On campaign financing, Public First Action, a non-profit that advocates for AI regulation, poured over $3 million into Wiener's campaign, and the group receives major funding from Anthropic. His opponents have attacked him over this, claiming he is too closely aligned with the AI industry. Wiener provided a written response after the interview: Public First endorsed him because he is the first legislator in the U.S. to introduce formal AI regulation, and the first to push such rules to passage and implementation. Those attacks deliberately ignore the fact that he pushed for strict AI regulation in direct opposition to big tech companies.

The political reality for his agenda is far from easy. He says he would be happy to join a dedicated AI select committee if Congress establishes one. Senator Bernie Sanders and several other lawmakers are potential allies on the issue of catastrophic AI risk. But there remains a gap between the Bay Area's widespread focus on AGI risk and the prevailing national sentiment. In local referendums held on Tuesday, 40 out of 41 anti-data-center measures passed, which appears to be a tailwind for AI regulation. Wiener however notes that most residents are pushing back against data centers popping up in their neighborhoods, driving up electricity bills and draining local water resources. The local backlash against data centers is not the same as a nationwide AI safety movement.

He says he is only "cautiously optimistic" that Congress will effectively regulate AI, pointing out that Congress has a poor track record of overseeing the tech industry. In September, Donald Trump called catastrophic AI risk a "hoax" and rejected the idea of cooperating with China on AI governance, a collaboration Wiener views as indispensable to avoiding the worst possible outcomes. Trump's unpredictability has created a fragile, unexpected window of opportunity for progress on the issue.

He notes that what we have seen so far are just red flags, not a full-blown catastrophe. He hopes the president does not need to wait for a disaster to take action, and if a disaster does end up being necessary to spur action, he hopes it will be as small as possible.

This article was originally published on MIT Technology Review, written by Grace Huckins: https://www.technologyreview.com/2026/11/09/1148182/nancy-pelosis-replacement-is-ready-to-regulate-ai/