WASHINGTON — In the final days before the midterm vote, social media videos believed to be produced with artificial intelligence and linked to Russia are misusing real people’s names and photographs to spread disinformation about two Democratic Senate candidates, according to the Associated Press. The targets are Sherrod Brown, running against Republican Senator Jon Husted in Ohio, and James Talarico, running against Republican Ken Paxton for a Senate seat in Texas. Both campaigns, and the people whose identities were used, say the material is fabricated — and researchers say the clips carry the hallmarks of a Russian-linked influence operation that ran similar campaigns two years ago.
The videos matter beyond two Senate races because of how they were built: not as crude text posts, but as believable video featuring real people, delivered through freshly created accounts at the moment voters are deciding which party controls Congress next year. Platforms removed the original posts, but copies kept circulating, showing how hard it is to contain a fake once it has been seeded in the last stretch of a campaign.
What the AI Disinformation Videos Show
The video targeting Brown, as described by the Associated Press, appears to include some AI manipulation and is presented in the style of a newscast. It uses pictures of Haya Akbik, who interned for Brown several years ago, and other former staff members. Akbik told the Associated Press that any claims attributed to her in the clip are “completely made up,” adding that it was her name and her reputation being used to make a false allegation. She said the episode showed the importance of Congress regulating what she called “AI slop on the internet,” calling the video deeply damaging.
The video targeting Talarico is framed differently: a direct-to-camera confessional from someone presented as a former student of Talarico, accusing him of sexual assault. Talarico’s campaign rejected the material outright. Spokesperson JT Ennis said in a statement that the most powerful people in the world know they cannot beat Talarico with the truth, so they are slandering him with AI deepfakes, foreign bots and conspiracy theories. The campaign’s position is that the person in the video is not who the account claims, and the accusation is false.
The social media company X took down the initial posts that shared the videos and suspended the users for violating its rules. But removal did not end circulation: copies could still be found online days later, some with hundreds of shares, a pattern researchers say is typical of late-campaign disinformation, where the goal is less to survive moderation than to reach enough voters before polling day.
How Experts Linked the AI Disinformation Videos to Storm-1516
The attribution to Russia rests on researcher analysis, not on a new official government finding, and that distinction matters. Clemson University professor and media forensics expert Darren Linvill said the new videos bear similar hallmarks to clips from a Russian network known as Storm-1516, which targeted former Democratic presidential candidate Kamala Harris and her running mate, Minnesota Governor Tim Walz, two years ago. Linvill found that accounts sharing the recent material had also shared former content from Storm-1516.
Graphika, a research firm that tracks influence operations, went further on the Talarico material, assessing with high confidence that Russia-aligned Storm-1516 produced the videos. According to Graphika’s analysis, a newly created X account seeded two very likely deepfake videos on October 20, depicting a supposed former student accusing Talarico of sexual assault, and the account appeared to impersonate a real Facebook user of the same name from Converse, Texas. The assessment rests on dissemination tactics and content characteristics, the firm said.
Forensic analysis supports the claim that the videos are not authentic. Hany Farid, a Dartmouth College professor who studies AI deepfakes, said his analyses showed the videos targeting Brown and Talarico were likely either generated or manipulated using the technology. He described the Brown video as a “fairly low-tech fake,” while the Talarico video is likely generated by AI and consistent with previous disinformation campaigns released close to Election Day.
What Is Proven and What Is Only Attributed
Three things are established on the record. First, the videos exist and were circulated as if they showed real people making real accusations. Second, forensic experts assess they were generated or manipulated with AI. Third, the people whose identities were used and the targeted campaigns say the claims are false — Akbik on the record to the Associated Press, and the Talarico campaign in its public statement.
What is not established is an official attribution. The Russian Embassy in Washington did not immediately respond to a request for comment, and in the past it has denied interfering in American elections. The United States Cybersecurity and Infrastructure Security Agency, which has publicly identified some videos as Russian disinformation in past election cycles, also did not respond to a request for comment. No American agency has, in the reporting reviewed here, issued a new formal finding naming Russia as the source of these specific videos. Senior Democrats Jim Himes and Mark Warner have said publicly that Russia is attempting to harm the Democratic Party and erode voter confidence ahead of the midterms, but that is a political assessment, not an intelligence finding.
Nor is it established how many voters believed the clips, whether they changed any votes, or how much of their apparent reach was authentic rather than amplified by bots and coordinated accounts. Researchers are careful on this point: network links and forensic indicators support the attribution assessments, but they do not measure electoral effect. Voters in Ohio and Texas will decide the races on Tuesday regardless of how the videos spread.
The Wider AI Election Environment in 2026
The Brown and Talarico videos are the most prominent examples of a much larger shift. The Wesleyan Media Project has tracked 220 midterm election advertisements generated, at least in part, using AI tools, with spending on AI-generated ads exceeding 137 million dollars earlier in October. Not all of that material is deceptive — campaigns openly use AI for production — but the volume makes it harder for voters to tell authentic campaign content from fabricated attacks, especially when a fake borrows a real person’s face and name.
That borrowing is what makes this case different from a conventional smear. A text claim can be denied and sourced; a video of a seemingly real person speaking feels like evidence, even when forensic analysis shows it is not. Experts say voters cannot rely on visual clues alone, because modern generation tools can reproduce lighting, lip movement and background detail well enough to pass a casual viewing. The practical advice from researchers is consistent: check whether a reputable outlet has verified the clip, look for the original account’s history, and treat anonymous, newly created accounts pushing explosive claims days before an election with suspicion.
Platform enforcement has also shown its limits. X removed the initial posts and suspended the accounts, yet copies persisted. Other platforms face the same problem: a video can be re-uploaded, clipped, screenshotted and re-captioned faster than moderators can chase every copy, and AI search tools can then pick up a debunked claim and repeat it as if it were still open.
What Happens Next Before and After Election Day
The immediate test is Tuesday’s vote. Ohio and Texas are both competitive Senate contests, and the videos arrived while early voting was already under way — the Talarico videos were seeded on October 20, a day after Texans began voting. Campaigns in both states have now put their denials on the record, and the Associated Press reporting gives voters a verified account of what the videos are and who assessed them.
The longer question is regulatory. Akbik’s call for Congress to regulate AI-generated material online reflects a debate that has produced little federal legislation so far, despite repeated warnings from researchers and election officials. States have moved faster in some cases, but a fake video can cross state lines instantly, and a federal framework would have to balance deception against free speech — including the right to use AI tools for satire, parody and legitimate campaigning.
For now, the pattern is set: a familiar influence network, updated with AI tools, using real people’s identities to deliver fabricated accusations at scale in the final days of a campaign. Whether voters, platforms and lawmakers treat this episode as a warning or as background noise will shape how the next election is fought — and how much of what voters see can be believed.
Sources
USN reviewed reporting from multiple outlets for this story. Read more from Associated Press, Associated Press via The Economic Times and Cyfluence Research.