Nancy Pelosi’s replacement is ready to regulate AI

Scott Wiener, the California state senator behind SB 1047 and SB 53, won Nancy Pelosi's open San Francisco House seat in a tight race. That puts the country's most experienced frontier-AI legislator into Congress at a moment when federal AI policy is wide open.

In his interview with MIT Technology Review, Wiener said he wants a strong federal regulator, whistleblower protections, testing and auditing requirements, a kill switch requirement, and aggressive oversight of the labs. He opposes broad federal preemption of state AI laws. He also said he would want to serve on a select committee on AI if Democrats create one. This is someone who has already passed AI law, not just talked about it.

We think Wiener's arrival is good for the quality of the debate, even where we'd push back on parts of his framing. He is one of the few elected officials who has actually taken a frontier AI bill through the full legislative grind, including a veto and a narrower second attempt that became law. That experience is scarce in Washington, and it should make federal proposals more workable.

His policy list is mostly sensible. Whistleblower protections, independent testing, auditing and a credible ability to stop a dangerous system are reasonable guardrails that don't block useful AI from shipping. We'd add that a federal regulator only helps if it's fast and technically competent; a slow one becomes a tax on smaller builders while big labs absorb the compliance cost.

Where we're more cautious is the urgency claim. Wiener cited METR's recent report, which has been widely read as evidence of recursive self-improvement at Anthropic, and said we could sleepwalk into a disaster. That interpretation is contested, and legislation written in a panic tends to age badly. The likelier path to durable rules is grounding them in measurable capabilities and incidents rather than a single report's reading.

The funding question will follow him. His campaign drew more than $3 million from Public First Action, a pro-regulation group that gets substantial Anthropic funding. Wiener calls the criticism cynical and points to his record of regulating over industry objections. That's a fair defense on the record. But when a leading lab funds the advocacy behind the rules that will govern it and its rivals, the public deserves transparency about how that influence works, and we'll keep asking.

The political gap may be the biggest obstacle. Voters approved 40 of 41 anti-data-center ballot measures last week, but Wiener himself reads that as local anger about electric bills, water and neighborhood impact, not concern about loss of control. Meanwhile the President called catastrophic AI risk a hoax in September. A Bay Area safety agenda will need to connect with kitchen-table concerns to go anywhere. Our read: the data center fight, not existential risk, is the doorway to federal AI policy.

SB 1047, authored by Wiener in 2024, would have been the country's first major frontier AI law before Gov. Gavin Newsom vetoed it. Its lighter successor, SB 53, was signed. Both have inspired proposed federal bills. Rep. Ro Khanna has proposed legislation to create a federal AI regulator, which Wiener praised.