The US midterm election campaign has seen a rise in political ads utilizing AI deepfakes that may mislead voters. Bruce Blakeman, a Republican candidate for governor of New York, released ads featuring AI-generated images of New York City Mayor Zohran Mamdani and Governor Kathy Hochul gardening together and discussing price increases for residents. Another ad depicts Hochul discussing toll and fare hikes that could make transportation unaffordable.
Neither Hochul nor Mamdani appeared in the ads, which were created using artificial intelligence. The use of AI-generated ads, which are clearly fake yet capture voter attention, has become prominent as the midterm elections approach on November 3.
Regulatory efforts have focused on labeling AI deepfakes that could mislead voters, but Scott Babwah Brennen, director of New York University’s Center on Technology Policy, described these ads as “slopaganda,” or identifiably fake ads. Political candidates and their supporters have spent over $80 million on more than 164 AI-based ads this election season, with Republican candidates and their backers responsible for 80 percent of these ads, according to the Wesleyan Media Project, a non-partisan election advertising tracker.
Blakeman defended his ads, stating they were clearly labeled as AI-generated, with disclaimers appearing for a few seconds at the beginning. He argued that they follow a long tradition of satire and parody in American politics. However, Brennen noted that slopaganda mixes low-quality AI “slop” with propaganda, potentially shaping perceptions of candidates’ character or ideology and impacting voter trust.
Other AI-generated ads included portrayals of Sherrod Brown, a Senate Democratic candidate from Ohio, as the last guest at a party, and James Talarico, a Senate Democratic candidate from Texas, depicted in a blue dress reminiscent of the character Maria from The Sound of Music. Mike Rogers, a Senate Republican candidate from Michigan, was shown as a superhero.
In San Francisco, Democratic California state Senator Scott Wiener faced backlash after releasing an AI chatbot that answered questions about his opponent, Connie Chan, parodying her record on real estate development and her background. The chatbot was criticized for being bigoted, leading Wiener to withdraw it and acknowledge that it “missed the mark.”
Deepfake ads have emerged in a political landscape that seeks to protect political speech, including parody. While no federal law regulates AI deepfake ads, 31 states have enacted laws requiring such ads to be labeled. Ilana Beller, director of State AI Governance and Policy at Public Citizen, stated that state laws attempt to balance regulation without infringing on the First Amendment, but emphasized the need for federal legislation.
Adam Rose, chief operating officer of Starling Lab, a research center on technology and human rights, noted that some AI modifications may not require labeling, such as color or voice corrections. He argued that candidates make judgment calls on their ads, which reflect on them. Regarding Wiener, Rose commented that releasing the chatbot demonstrated poor judgment.
Wiener is running for a congressional seat in San Francisco, a city central to the AI industry, with major companies like OpenAI and Anthropic based there. The representative will play a crucial role as legislation surrounding AI develops, especially as polls indicate that many Americans are more concerned than excited about the technology.
Ian Krager, a spokesperson for Connie Chan, stated that the race represents a pivotal moment for the country, highlighting the potential for America to lead in AI development without imposing costs on working people. He asserted that Chan supports working families, while Wiener aligns with AI companies, referencing the chatbot incident.
As a state senator, Wiener drafted a California law requiring AI companies to disclose security incidents involving autonomous systems created by large AI models. A less stringent version of this law was enacted last year. Wiener’s office did not respond to a request for comment.
Looking ahead to the midterms and beyond, Beller stressed the importance of making AI ad labeling clear and understandable, akin to food labels, to help voters evaluate them. She noted the need for increased awareness about AI deepfake ads, particularly among younger voters and seniors.
In New York, the Democratic Party filed a complaint with the state board of elections, claiming that Blakeman’s ads violate disclosure requirements and mislead voters. Blakeman countered that many find the ads humorous and appreciate the satire.
Brennen remarked that the ads aim to attract attention, as candidates recognize the political power in doing so. As the use of AI deepfake ads continues to grow, some candidates have publicly stated they will refrain from using them, with Beller noting that this decision seems to resonate positively with voters.