State Sen. Scott Wiener launched an AI chatbot impersonating his congressional opponent Connie Chan, and when it generated a response mocking her accent, former House Speaker Nancy Pelosi led a coalition calling the tool racist and sexist.
The chatbot, hosted at the domain ConnieChan.ai, went live earlier this week as Wiener and Chan battle for the San Francisco congressional seat Pelosi once held. Wiener's campaign says the tool was labeled as parody and built to spotlight Chan's record on housing, transit, and education. But one prompt produced a response that read: "I'm afraid my San Francisco accent is so thick I can barely understand myself." Chan emigrated from Hong Kong to San Francisco at age 13, and the accent line set off a firestorm among Democratic officials who back her candidacy, The San Francisco Standard reported.
Pelosi and four members of Congress, Reps. Zoe Lofgren, Grace Meng, Jill Tokuda, and Judy Chu, joined San Francisco Supervisors Jackie Fielder and Chyanne Chen in signing an open letter released Thursday condemning the chatbot. Every signatory has endorsed Chan.
The letter did not mince words. Pelosi and her co-signatories called the chatbot "a gross misuse of technology" that "reflects a disturbing willingness to erase women's agency." They framed the issue as both a gender and a racial offense:
"Women have spent generations fighting to have our own voices heard in the halls of power. We should not have to fight against artificial intelligence putting words in our mouths."
The letter went further, arguing that targeting a woman of color compounds the damage. The signatories wrote that the chatbot's output echoed "harmful racial tropes" and likened it to "Trump birtherism", a comparison that ties Wiener, a progressive state senator, to language his own party has spent years denouncing.
California Lt. Gov. Eleni Kounalakis piled on separately, posting on X that the campaign should pull the tool down entirely:
"The Wiener campaign should take down this chatbot. San Franciscans deserve campaigns that are honest, respectful, and focused on earning voters' trust, not fabricating someone else's words."
So a Democratic state senator built an AI tool to put fabricated words in a rival's mouth, and a bipartisan chorus, led by the most powerful Democrat in San Francisco history, is calling it a racist stunt. That is the fight Wiener now has on his hands as Pelosi has thrown her endorsement behind Chan in the race to succeed her.
Wiener did not apologize. His campaign issued a statement defending the chatbot as clearly labeled parody focused on policy:
"Our campaign created this chatbot, and clearly labeled it as a parody, to highlight Supervisor Chan's abysmal record on affordable housing, public transportation, education, and San Francisco's economic recovery. These are critical issues for San Francisco's future, and the chatbot was designed to focus only on her record."
Then he pivoted to offense. Wiener pointed out that Pelosi herself played a key role in helping defeat SB 1047, his 2024 AI regulation bill that passed the California Legislature before Gov. Gavin Newsom vetoed it. Pelosi had publicly called the bill "more harmful than helpful." Newsom's veto message said the legislation applied "stringent standards" to even the most basic AI functions.
The irony Wiener is pointing at is real: Pelosi helped sink a bill that would have imposed tighter rules on AI tools, and now she is condemning a specific AI tool as dangerous. Whether that undercuts her letter or simply proves her selective about which AI uses bother her is a question voters in the district will have to sort out. Pelosi's political influence in California has already shown signs of erosion in recent cycles.
Wiener's campaign offered one data point to bolster its defense: it claimed users submitted roughly 109 prompts asking the chatbot about Chan's accent, race, gender, appearance, or citizenship. The implication was that provocateurs, not the chatbot's design, drove the offensive output.
But when The San Francisco Standard asked for the backend data to independently verify that number, the Wiener campaign declined. That refusal leaves the 109 figure as an unverified campaign claim, not an established fact. A campaign that builds an AI tool to impersonate a rival and then refuses to open its data to scrutiny is asking for trust it has not earned.
Chan's campaign spokesperson, Ian Krager, rejected Wiener's framing entirely:
"Scott Wiener can try to change the subject, but he can't change the facts. He built a chatbot with his Big AI donors to impersonate Connie Chan and is now attacking women who raised legitimate concerns about its racist and sexist responses."
Krager's reference to "Big AI donors" raises its own unanswered question, he did not name the donors, and the source material does not identify them. But the accusation plants a seed: that Wiener's AI industry ties shaped the tool's creation.
This controversy lands at a precarious moment for Wiener. After the June primary, he led Chan by 11 points. Recent polls obtained by Mission Local show that gap has collapsed to a statistical tie. A race that once looked comfortable for Wiener is now a dead heat, and the chatbot flap gives his opponents fresh ammunition to define him as reckless.
Chan's base sits primarily in the Richmond district on San Francisco's west side. The chatbot had targeted her record on specific local issues, opposing a bus rapid transit corridor on Geary Boulevard over parking concerns, and opposing higher Bay Bridge tolls to fund transit projects. Those are legitimate policy contrasts. But the accent line turned a policy attack into a racial flashpoint, and Wiener's campaign has struggled to separate the two.
The broader San Francisco congressional race has already been a bruising affair. Former AOC chief of staff Saikat Chakrabarti poured $10 million of his own money into the primary and failed to win a single precinct. Pelosi herself publicly attacked Chakrabarti during the race, making clear she intended to control the succession. Now her endorsement and her letter make Chan the establishment favorite, and Wiener the target.
What makes this episode worth watching is not the chatbot itself. Parody accounts, opposition-research websites, and satirical impersonations are old campaign tactics dressed up in new technology. The question is whether the racism charge sticks, and whether the officials making it have standing to do so.
Every signatory of the Pelosi letter endorsed Chan. That does not make their objections wrong, but it does make them interested parties. Pelosi has shown no reluctance to use sharp tactics against candidates she opposes in this same race. The letter reads less like a principled stand on AI ethics and more like a campaign surrogate operation with a civil-rights wrapper.
Wiener authored the most significant AI regulation bill California's legislature has produced, and Pelosi helped ensure it never became law. Now Pelosi is leveraging an AI controversy to damage the man who wrote the regulation she opposed. That is not hypocrisy in the technical sense, she can oppose a specific bill and still oppose a specific chatbot. But it does reveal how selectively San Francisco's Democratic establishment applies its concern about AI harms: urgently when it helps a preferred candidate, and not at all when a regulatory framework is on the table.
Voters in San Francisco's congressional district are watching two Democrats accuse each other of bad faith while competing for a seat that has been in Democratic hands for decades. Neither side is asking whether an AI chatbot that impersonates a real person and generates fabricated quotes should be legal. They are asking whether it helps or hurts their campaign.
That tells you everything about the seriousness of the AI ethics conversation in Democratic politics right now: it lasts exactly as long as it is useful.
