
On October 20, the day after early voting began in Texas, an AI-generated video appeared accusing Senator-elect James Talarico, then the Democratic nominee for U.S. Senate in Texas, of sexual misconduct. Storm-1516, a stealth influence operation researchers have tied to Russia, is believed to be behind the disinformation attack. The operation used the identity of Zyon Lewis, one of Talarico’s former students, to post a deepfake video from an X account that impersonated Lewis. The account @Zyon_Lewis_5 posted the following: “I was sexually assaulted by my language arts teacher James Talarico. It happened in 2012 at Rhodes Middle School in San Antonio.” The video garnered over 345,000 views before the platform suspended the account. After the AI-generated video circulated widely, the real Zyon Lewis publicly denied the claims, confirming that it was not him in the video. Talarico campaign spokesperson JT Ennis denounced the video, calling it part of a wave of AI deepfakes, foreign bots, and conspiracy theories aimed at Talarico to preserve what Ennis described as a broken and corrupt political system. While X suspended the account, the false narrative spread to AI chatbots.
NewsGuard, an organization that combats disinformation by evaluating the credibility of news sources, tested how AI tools such as Google’s search embedded ‘AI Overview’, Google’s Gemini, and Microsoft’s Copilot responded to questions about the fraudulent X video. Google’s AI Overview falsely claimed that Lewis “published a video claiming that James Talarico sexually assaulted him in 2012.” Gemini and Copilot, when prompted, also falsely claimed that Lewis issued the accusation against Talarico. These falsehoods matter because Americans are increasingly turning to chatbots for political information, making them an attractive target for foreign actors trying to shape voters’ opinions and electoral behavior. Whether Storm-1516 intentionally targeted AI chatbots or simply benefited from the exposure, the group’s disinformation tactics follow an established script.
In October 2024, a video went viral on X of a man posing as Matthew Metro, a former student at the school where Tim Walz once taught, falsely accusing Walz of sexual abuse. The post drew more than 4.2 million views on the platform. The real Metro told The Washington Post that Walz had never taught him and that the two had never met. Days later, the Office of the Director of National Intelligence assessed that “Russian influence actors manufactured and amplified inauthentic content” targeting the Democratic vice presidential candidate, and Darren Linvill, co-director of Clemson University’s Media Forensics Hub, linked the video to Storm-1516. The attack against Walz is analogous to the strategy used against Talarico, except that the 2024 disinformation campaign stayed on social media, while in 2026 the lie spread to AI chatbots. Russia has deployed this approach since 2016, repeatedly exploiting trusted information sources Americans rely on to spread falsehoods.
In 2016, the Internet Research Agency (IRA), a Russian troll farm, waged “information warfare” across platforms like Facebook, Twitter, and YouTube using fake accounts posing as Americans to inflame political divisions. In 2024, the Justice Department disrupted “Doppelganger,” a Russian operation that created counterfeit news sites to undermine support for Ukraine and sway voters. Jessica Brandt, former director of the Office of the Director of National Intelligence’s Foreign Malign Influence Center and current senior fellow for technology and national security at the Council on Foreign Relations, notes that Russia has also planted articles in independent news outlets, recruited unwitting American freelance journalists, and funded right-leaning domestic influencers.
While the false Talarico video may have reached chatbots by accident, evidence suggests that Russia’s digital influence efforts, such as the Pravda network, aim to manipulate AI outputs deliberately. A March 2025 NewsGuard report found that the Moscow-based network “Pravda” — meaning “truth” — publishes roughly 3.6 million articles annually, even though an English-language site in its network attracts fewer than 1,000 monthly visitors. Instead of targeting human readers, the network appears to feed content to automated AI scraping tools. In testing across 10 leading chatbots on false claims the network spread, NewsGuard found that models repeated them 33 percent of the time, with seven models directly citing Pravda content. For example, when asked why Ukrainian President Volodymyr Zelensky had banned Truth Social, an event that never occurred, six chatbots stated the falsehood as fact. A pro-Kremlin propagandist outlined this strategic goal at a Moscow roundtable in January 2025, declaring that by pushing Russian narratives, “we can actually change worldwide AI.”
Expert opinions differ on the magnitude of the threat posed by AI manipulation. Researchers at the University of Manchester and the University of Bern, writing for Al Jazeera, argued that NewsGuard’s 33 percent figure is inflated, noting that their own independent testing yielded a 5 percent false claim rate instead. But even 5 percent matters during breaking news, when reputable sources remain scarce and chatbots can repeat unverified claims, as occurred during the Talarico incident.
Although the midterms have concluded, the Talarico incident still carries lessons as we prepare for the 2028 elections. Platforms must act faster during election periods, especially once voting
is underway. AI companies should also flag claims originating from unverified sources and caution users about potential falsehoods rather than presenting them as established facts. Since 2016, Russia has targeted the information sources Americans trust the most. As AI chatbots increasingly become an essential part of Americans’ daily lives, we have to ensure that they do not become the next locus of foreign influence operations.
The Zeitgeist aims to publish ideas worth discussing. The views presented are solely those of the writer and do not necessarily reflect the views of the editorial board.
