As election season enters the final three months before voters hit the polls, Scott Wiener’s campaign has whipped up a novel tactic in San Francisco politics: an AI chatbot that satirizes and disparages Connie Chan’s record as a supervisor in the city.
The site, conniechan.ai, allows users to ask the “Connie” bot questions or click on several key issues, such as Chan’s purported opposition to housing development, public transit funding, and law enforcement surveillance tools.
“Let’s do nothing together,” the site’s header states.
It continues with an explainer: “I’m Connie, the world’s first AI trained to say ‘no’ to anything you ask. Bring me an idea and I’ll explain why it can’t, shouldn’t, and won’t happen.”
“No lies were told in the making of this AI chatbot,” Joe Arellano, spokesperson for the Wiener campaign, told Gazetteer SF in an email. “This is 100% Connie Chan’s record of blocking all progress, packaged in an easy to use way for voters.”
The Wiener campaign was in part inspired to create the chatbot in order to “lean into the AI craze overtaking the city,” Arellano added.
The chatbot is a kind of online interactive attack ad. Jason McDaniel, a professor of political science at San Francisco State University, told Gazetteer that this is the first time he has observed an AI chatbot being used in city political campaigns.
The novelty of the chatbot may help the campaign in two ways, he said.
“First, it may be more memorable to people who are exposed to it [or] experience it. Attack ads have very little effect on voters, especially attack mailers. But some research shows that they are more memorable to people, even though they don’t have much if any effect at all,” McDaniel wrote in a text message. “Second, the novelty of an attack ad like this one may help the campaign by driving indirect attention through media coverage.”
The Wiener campaign did not answer Gazetteer’s questions about when it launched the site and how much it has spent to create and operate the chatbot using Anthropic, Amazon Web Services, and Google Analytics, which are listed on the site’s privacy policy. A lookup of the site’s domain name shows that the “conniechan.ai” domain was purchased on June 16.
The campaign also did not address a question about potential data privacy concerns by users. The chatbot’s policy states conversations are deleted after 90 days but that “whatever you type is what we keep. Please do not type anything into Connie that you would not want us to have — your own sensitive information, or anyone else’s.”
A “small number” of authorized staff review conversations to fix bugs and to understand “what people ask about” so that the campaign can “write better content,” the policy states.
The guidance to not type any sensitive or personal information is not mentioned on the main page of the site or in the chatbot conversation window. The privacy policy emphasizes that conversations are not used to train the campaign’s AI model, and that Anthropic has given a “commitment” to not use data from the website for AI training under its own privacy policy.
Ian Krager, spokesperson for the Chan campaign, told Gazetteer that Wiener is “desperately resorting to gimmicks” because Chan’s message to voters has narrowed the gap; recent polling suggests that Chan is within five percentage points of Wiener.
“If there was ever any doubt that Scott Wiener is bought and paid for by Big AI, he’s not even trying to hide it anymore,” Krager said. “We know Scott Wiener is desperate for attention, but it’s never a good idea for a man to put words in a woman’s mouth — especially when they are falsities fraught with racist and sexist undertones.”
Krager added that Wiener has “cozied up to Big AI and corporate money,” and that “voters deserve better than the tired and trite slop Scott Wiener is offering.”
Wiener has called for a number of tech regulations, including national legislation for net neutrality and AI safety measures that he has pushed on the state level. The AI industry has supported his candidacy, including $500,000 from an Anthropic-funded PAC called Public First. Chan has pledged to not take corporate PAC dollars, and told Mission Local in April that she is “skeptical” about AI and would work in Congress to limit job losses from AI expansion.
“Scott wrote the landmark AI safety laws serving as the model for the rest of the country. He’s been thinking about workers and safety and delivering on these issues for years,” said Arellano, Wiener’s spokesperson. “He won’t be learning on the job like Supervisor Chan.”
Editor's note: This story has been updated to clarify that a PAC supporting Scott Wiener took money from an Anthropic-funded PAC. His campaign did not take any money directly from the company or its PAC.
