AI Campaign Ads Test Oregon's Deepfake Law as Midterm Elections Heat Up

AI campaign ads put Oregon's deepfake law to the test

Why it matters: AI-generated attack ads and campaign content are rampant leading up to the midterms, with candidates across the political spectrum using the technology. State of play: Oregon Senate Bill 1571, passed in 2024, requires candidates to disclose when they use AI to create campaign ads. * Minnesota and Texas prohibit political deepfakes for a certain number of days before an election. Maryland's ban is year-round. * Yes, but: Similar laws in California and Hawaii were struck down in court on First Amendment grounds, according to the National Conference of State Legislatures. Zoom in: Oregon's AI election law is getting its first test this year after complaints about ads run by Jonathan Lockwood, who unsuccessfully ran in the 5th Congressional District primary to unseat U.S. Rep. Janelle Bynum, OPB reported. * One ad showed Bynum and Gov. Tina Kotek with glowing red eyes chasing a group of scared children. * Another showed Bynum in a McDonald's uniform handing a drag queen keys to a girl's locker room and saying "Don't let me down." * The Secretary of State's office initiated an investigation in July following a complaint. Lockwood wrote to state officials that the ads were so outlandish that no one could mistake them for the real thing. * The Secretary of State's office told Axios the investigation is ongoing. * If he's found to have broken the law, Lockwood could face prosecution and a fine of up to $10,000. The big picture: At the federal level, lawmakers have yet to establish a baseline standard for election AI deepfakes. * Congress did attempt to tackle nonconsensual intimate deepfakes -- an issue that has collided with election ads targeting women lawmakers. * But implementation has been difficult, advocates say, citing inconsistent processes for removing posts and predicting litigation will be needed for strong enforcement. * House Democrats say they plan to push legislation next year targeting election deepfakes if they regain power. The bottom line: Proposals to tackle AI in elections have been floated in Congress for years, but Oregon has proven ahead of the curve, at least so far.

AI reshapes traditional campaign strategies

Artificial intelligence is reshaping the campaign trail, as candidates up and down the ballot turn to the technology for deepfake attack ads, viral social media posts and self-promotion. The embrace of AI as a political tool varies across political lines and races, drawing debate over whether deepfake content -- particularly realistic videos or audio -- could mislead voters or disrupt the democratic process. Proponents argue AI-generated content just marks a new chapter for campaign advertising, building upon decades of political cartoons and graphics for a cheaper price. The conversation was amplified last week, when the social conservative group American Principles Project (APP) released a digital ad in support of Texas Attorney General Ken Paxton (R) depicting his Senate race opponent, Texas state Rep. James Talarico (D), as a doctor performing gender-reassignment surgery on a little girl. "We took the approach of using AI in the same way that you would use political cartoons to articulate a point or a message about the consequences of someone's voting behavior or their actions while in office," Terry Schilling, the president of APP, told The Hill on Monday. The 30-second digital ad resembles the 1964 "Daisy" ad for former President Lyndon B. Johnson, which played on fears of the Cold War nuclear race, and is part of a larger Republican movement targeting Talarico over his support of transgender people. The ad, released last Thursday, features a girl counting flowers in a field as the narrator says "she dreams of growing up, falling in love, becoming a mother." The girl is then in a hospital bed counting down as a deepfake Talarico stands above her, and a narrator says, "The trans industry preys on confused children. Irreversible puberty blockers, cross-sex hormones, sex-change surgeries ... this November, stop them." Schilling said the group used AI to depict what Talarico's policies could "look like in reality," but JT Ennis, a spokesperson for the Talarico campaign, told The Hill the ad is "flat-out lying" and that the Texas Democrat does not support gender-transition surgery for minors. "This ad exemplifies the rot at the core of our broken, corrupt politics: Ken Paxton and the billionaires who prop him up are using shameful scare tactics and disgustingly using children as political props to lie about James because they have no solutions to lower the price of gas, groceries, and healthcare," Ennis wrote. It is the latest of dozens of AI-generated attacks across the country this cycle and follows another Paxton ad in March showing an AI-generated version of Talarico in a dress singing about transgender children. Other viral attack ads have featured former Los Angeles mayoral candidate and reality television star Spencer Pratt, who was depicted in a May ad as a hero in a Gotham-influenced, dystopian hellscape, while incumbent Los Angeles Mayor Karen Bass (D) was shown as the Joker. Pratt lost the GOP primary nomination in June. The ad was released by filmmaker Charles Curran of Menace Studios with no official affiliation with Pratt, but the then-candidate posted the video, which racked up more than 4 million views and thousands of reposts. Bass at the time accused the ads of "taking on a violent trend" that was "very scary." In New York, Republican gubernatorial candidate Bruce Blakeman is facing an ethics complaint over his campaign's use of AI in ads against New York Gov. Kathy Hochul (D). The chair of the New York State Democratic Committee argued the ads are an attempt to "distort reality." And during the GOP primary for Tennessee's 6th Congressional District, state Rep. Johnny Garrett, who clinched the nomination last week, and former Rep. Van Hilleary both used AI in campaign ads against each other. In doing so, they may have violated a new Tennessee law requiring certain AI disclosures. The use of AI for campaign purposes spreads across party lines, with high-profile Democrats such as California Gov. Gavin Newsom and former New York City Mayor Eric Adams also utilizing the technology to attack political enemies. As AI's image, audio and video capabilities grow, the content is becoming more realistic and accessible to not only campaigns but also everyday users online. Schilling said the latest anti-Talarico ad was the first from the PAC using AI, stating past AI-generated ads looked "cheap." "The technology is only getting better, I really could see a day where you wouldn't even be able to tell at all," Schilling said, adding it was "much cheaper" than using an entire film crew. While some political organizations are embracing the advancing technology, others are concerned about the spread of misinformation, even if a reasonable viewer could understand it is not real. Research by Daniel Schiff, an assistant professor of technology policy who studies deepfakes at Purdue University, found that while most can tell whether something is a deepfake or satire, this does not prevent the content from potentially influencing voters. "Even if they know it's a deepfake ... that doesn't necessarily correlate to their beliefs," Schiff said. "So we'll ask people, 'Do you think this is literally true?' And then we'll say, 'Even if it's not literally true, does it represent something that feels true about the world?' And they'll say, 'Yes.'" "In many ways, the deepfakes can provide a sort of cover for different types of satire and propaganda and misinformation to still reach people and still influence people," Schiff said. "The misinformation problem is much deeper." Outside of satire, concerns are also growing around "hyper-realistic" videos and audio and the timing of their release. "The more deeply problematic ones" will be those released a month or week before an election, "like some cellphone call that happened or an alleged argument that happens at a campaign event," said J.B. Branch, the director of federal AI governance and technology policy at the progressive consumer rights watchdog Public Citizen. Groups like Public Citizen are pushing for regulations at the state level to require official campaign ads to clearly disclose when AI-generated content is used. More than two dozen states have passed legislation to this end, but managing compliance becomes tricky when any digital user can create and share potentially misleading ads. Unlike traditional ads, which can require professional filmmakers, the AI tools are being put "in everyone's hands," Schiff said. "Capabilities are improving, and, perhaps even more importantly, they're becoming naturalized in everyday workflows." Efforts to ensure more transparency about AI in election ads have been discussed at the federal level, but Congress has stalled on most AI-related legislation. Rep. Yvette Clarke (D-N.Y.), who first introduced legislation in 2023 to require such AI disclosures, said Congress is "negligent" over risks of deepfakes and AI as a whole. "It's a vulnerability that we have not addressed," Clarke told The Hill in an interview Tuesday, adding, "There may be a motive there, but at the end of the day, it's malpractice as legislators not to establish rules of the road." Clarke pointed to the White House's prevalent use of AI, "Once the White House is doing it, where do you go from there?" Still, attempts to regulate AI in election ads could face an uphill battle as most content falls under protected free speech. "Generative AI does not change the fundamental speech at hand. In fact, this ease of use and lower costs can make it easier for candidates to be creative or communicate with constituents," David Inserra, a fellow for free expression and technology at the libertarian think tank Cato Institute, wrote in a blog.