Scott Wiener and Connie Chan both want to rein in AI. They disagree on how to do that. 

Several months ago, campaign door-knockers in San Francisco’s congressional race to replace Nancy Pelosi might have only fielded general questions about their bosses’ stance on AI.

But after Anthropic researcher Jacob Coxon’s Sept. 8 public flameout triggered widespread alarm about AI firms “gambling with our lives,” campaign footsoldiers are boning up for somewhat different conversations. 

Volunteers and organizers with state Sen. Scott Wiener’s campaign are training themselves on how to give voters a crash course on AI safety — starting with translating industry jargon into terms voters can grasp, according to two campaign workers.

Words like “guardrails” or “kill switches” are now to be avoided for campaign canvassers, the workers said. Both connote danger. Instead, better to use “safety controls” and “emergency shutdowns.” 

Still, AI hasn’t moved to the front of either campaign’s ground game, according to workers from both campaigns. The issue has taken the backseat to more urgent questions about affordability between voters and canvassers for Wiener’s rival, Connie Chan, according to one worker for the District 1 supervisor’s congressional campaign.

It’s just one measure of how AI safety, which has lagged behind cost-of-living concerns in San Francisco, has become a growing focus in November, affecting everything from the language canvassers use to closer scrutiny of campaign donations, a candidate’s expertise and their policy statements. 

Doorknockers, the candidates, and their positions

“We went back to school with ourselves,” one of the two Wiener campaign workers said. “We’ve had to do training with field staff to take these terms and make them more accessible to the public.” 

It’s about meeting voters where they are, one worker said. “Our canvassers say — after that Jacob Coxon tweet — people have gone from asking, ‘What is your stance on AI?’ to ‘How will you regulate it?’” 

The issue has “taken on a life of its own,” the worker added. 

Chan has — along with more than a dozen lawmakers across the country — signed the “AI Pact,” in which the candidate committed to opposing tax breaks and subsidies for data centers, mandating safety reviews for major AI products before release, supporting localities’ rights to regulate AI, and other stances. 

Wiener, in turn, has publicly pointed to his record of authoring first-in-the-nation AI safety bills, one of which, SB 1047, was vetoed by Gov. Gavin Newsom after critics — including Pelosi, who is backing Chan in the race — dismissed the law as based on “hypothetical” concerns. 

Some of the bill’s critics have since switched their stance on Wiener’s original bill amid consensus from AI safety advocates that Newsom’s veto was a mistake. 

“The risk of an AI-driven catastrophe has rocketed up in the public consciousness. People are rightly concerned that AI is being built without guardrails or oversight, and they want to see someone with a plan to tackle those issues on day one,” Wiener said. “I began working on these issues in 2023, and I have the knowledge and the skills to lead the fight to protect the public from AI risks on day one.”

Chan said her approach regulates AI based on its capabilities, how it is used and whom it could harm.

“The defining question is whether AI will improve life for working people or become another tool for billionaires and corporations to eliminate jobs, raise costs and concentrate wealth and power,” Chan said. “Scott’s so-called signature law focuses on only the largest frontier models and the aftermath of catastrophic circumstances.”

U.S. lawmakers — some of whom severely lack AI literacy — are now scrambling to come to some bipartisan agreement on comprehensive federal AI regulation where none currently exists.

Both Wiener and Chan agree that AI will displace large numbers of workers, needs strict safeguards and monitoring, and that Congress has been slow to act. Here’s the split: Chan supports a data center moratorium, while Wiener supports data centers, but with taxes and strict conditions. 

While both tout platforms on strict federal oversight, independent auditing, and the protection of workers, Chan’s message leans heavier toward protecting working people through policies banning the automation of firing and disciplinary decisions and requiring companies report AI-driven job losses; Wiener’s message focuses more on federal safety and transparency standards building on California’s SB 53, his AI safety measure that earned Newsom’s approval after watering it down from an earlier, vetoed law that would have required AI firms create shutdown capabilities for dangerous models and third party auditing. Newsom has, in recent days, reversed course with executive orders requiring both of those aspects found in Wiener’s original bill.

And while Wiener argues for a new, civilian-led federal body to keep AI development leashed, Chan proposes strengthening existing agencies, including the Federal Trade Commission, Securities and Exchange Commission and Consumer Financial Protection Bureau.

“Scott is leaning into a much more technocratic, ‘I tried to regulate AI with bills that got passed and killed for the past couple of years’ approach, which is his approach to most issues,” said Jane Natoli, the former Bay Area director of YIMBY Action and a longtime Wiener supporter. “Connie is leaning into a much more populist element of it — making sure regular people aren’t left behind by AI.”

AI has also taken the race down ugly paths. After public pressure on Aug. 7, Wiener took down a chatbot his campaign designed to impersonate Chan and surface negative talking points (it responded to users by saying “no” to much of what it was asked). Wiener ultimately admitted the chatbot missed the mark.

Could campaign support become an issue? 

Wiener, for his efforts on state AI regulation, has strong ties to the AI community, including organizations like the Center for AI Safety, which worked with Wiener on SB 1047. One former OpenAI employee who now advocates for AI safety said Wiener is one of just a few lawmakers in the country who have thought in detail about these issues for years.

“Wiener has more expertise and experience in the matter versus Chan,” the former employee told Mission Local. 

Chan and her supporters said Wiener’s regulatory messaging is undercut by his own campaign support. Wiener has seen hundreds of thousands of dollars in contributions from workers at OpenAI and Anthropic. He also saw $500,000 in support from an Anthropic-backed Super PAC called Public First.  

“Scott opposes a national data center moratorium, let the industry shape his legislation, used AI to impersonate me and now benefits from $500,000 committed by an Anthropic-funded PAC,” Chan told Mission Local. “AI companies cannot be allowed to write their own rules — or bankroll the politicians responsible for enforcing them.”

Nadia Rahman, a progressive political strategist who worked on Saikat Chakrabarti’s unsuccessful primary campaign against Wiener and Chan, said AI spending in support of Wiener will “become a real issue.” 

“We’re going to see Chan’s campaign talking more about it, you’ll see it come up in debates, highlighted not only by supervisor Chan’s campaign but by organizations supporting her,” said Rahman, who now supports Chan. “It’s going to become a real hindrance to him.”

Wiener countered that many of those donations came from Open AI and Anthropic workers who split from their bosses and supported AI safety. Some in Wiener’s camp have questioned whether Chan will support Project Blueprint, a new AI lobbying initiative launched by Silicon Valley venture capitalist and political mega-donor Ron Conway, given Pelosi — whose endorsement of Chan has become signature to her campaign — has come out in support of it. Chan and Pelosi don’t align on everything, however. Chan supports “Medicare for All” while Pelosi opposes it.

Chan didn’t take a for or against stance on Project Blueprint when asked, though she voiced skepticism through a spokesperson.

“Any AI policy initiative backed by a billionaire investor must earn the public’s trust. Workers and their unions, antitrust enforcers, and financial regulators need to be at the center of this conversation and have a seat at the table,” Chan’s spokesperson said. “Supervisor Chan is glad Project Blueprint acknowledges the risks to jobs and communities, but she’ll judge it by whether it supports enforceable rules that put working people ahead of the corporations building AI.”

Wiener said he opposes Project Blueprint.

“Big Tech and its investors should not be allowed to write their own regulations for AI,” he said.

Natoli expects that for most voters — including those in the birthplace of AI — the doomer-ism takes second to people’s immediate reality.

“Even in San Francisco, it feels like a pretty nascent issue for voters,” Natoli said. “A lot of people haven’t worked out how they feel, while some folks are telling us it’s going to kill us all.”