Federal intelligence officials have concluded that Russia has activated a covert digital influence campaign aimed at the United States’ upcoming November midterm elections, using fabricated celebrity videos to mimic trusted news broadcasts and inflame the country’s deepest political divisions. The operation, described in secret intelligence assessments and reported by American officials to The New York Times, is not intended to support any single party, according to U.S. officials. Instead, its purpose is to widen polarization, undermine confidence in democratic institutions, and generate chaos. The classified assessments indicate that operatives linked to the Kremlin have already begun seeding artificial intelligence-generated clips across social media platforms. These videos falsely show Hollywood celebrities commenting on competitive Senate races and are edited with logos, graphics, and production cues to make them appear as though they were aired by respected outlets such as CNN, the BBC, or The New York Times. In reality, the materials are entirely fabricated. The development signals a troubling evolution in foreign interference, as synthetic media and generative AI dramatically lower the cost and difficulty of producing convincing disinformation. While officials describe the current campaign as less extensive and less coordinated than previous Russian attempts to influence American elections, they emphasize that Moscow has carried out online influence operations during every U.S. federal election cycle since 2016. One official said the overarching goal is not necessarily to determine who wins, but to make Americans distrust what they see and read online and to intensify hostility across partisan lines. The NYT report relies on internal assessments and interviews with officials who are actively tracking the effort. These officials are concerned that the operation will accelerate as Election Day approaches, and that the use of deepfakes will make it harder for ordinary voters to distinguish legitimate journalism from malicious fiction. The first wave of material may have been relatively limited, but researchers who study Kremlin-backed behavior say it could represent the opening phase of a much broader influence campaign. The videos were deliberately designed to exploit both the recognition of well-known actors and the visual authority of established news brands, combining familiar faces with trusted logos in a way that makes disinformation more shareable and more difficult to debunk before it goes viral.
The specific videos identified in recent weeks have focused on Senate races, according to American officials and independent researchers tracking Russian influence activity. In the most prominent example, an AI-manipulated clip featured actress Jamie Lee Curtis apparently criticizing Democratic candidates over their policies toward transgender people. The video was constructed to look like a legitimate news segment, complete with the branding of a major network. Curtis’s spokesperson, Heidi Schaeffer, dismissed the clip without hesitation, telling reporters that there was “not a drop of truth” in it. Despite that denial, the post had been viewed nearly 200,000 times on X by the end of the day it appeared. Researchers say the videos were distributed in two waves during the past week on X and Bluesky, the decentralized social media platform that has attracted many users seeking an alternative to X. The posts came from a network previously linked to Kremlin troll activity and contained fabricated accusations aimed at Democratic candidates. According to the materials, Democrats were accused of corruption, antisemitism, and supporting the rights of the LGBTQ+ community. The content appeared intended to inflame divisions within broad Democratic constituencies, potentially alienating certain voter groups and encouraging internal conflict rather than simply attacking candidates from the outside. Although U.S. officials maintain that the operation aims to sow chaos rather than to elect a particular party, the material identified so far has largely targeted Democratic candidates in closely contested states. This pattern fits a familiar Russian playbook: amplify existing social grievances, surround them with false context, and encourage deep distrust of the political establishment. Independent analysts at Graphika and other organizations flagged the videos as possible products of a coordinated influence network. The production quality—including seamless network bugs, lower-third graphics, and newsroom-style transitions—suggests that the creators studied American television news formats carefully. At the same time, the videos contained telltale signs of synthetic generation, including subtle inconsistencies in facial movement and unnatural audio intonation. Researchers warn, however, that such flaws are becoming increasingly difficult for ordinary users to detect as the technology improves.
Jean Le Roux, a senior researcher at the social media analysis company Graphika, described the campaign as the beginning of a much larger effort. “We certainly expect this to gain momentum. I don’t think they will stop. I believe we are witnessing the beginning of a campaign focused on the U.S. elections,” he said. Le Roux and his colleagues noticed that U.S.-focused content began appearing shortly after election-related influence operations in Germany and Sweden, a sequence that researchers have observed in previous cycles. The networks suspected in this operation are known by various labels in open-source intelligence, including Operation Overload and Matryoshka. These names are used to describe clusters of accounts and bots that have remained active for years and frequently adopt the appearance of legitimate news organizations. They create fake outlet websites, impersonate real journalists, and use social media personas to lend authority to rumors and manipulated media. By branding disinformation with the names of trusted news outlets, the operatives gain a level of credibility that anonymous accounts or ordinary political memes cannot achieve. Russia’s methods have evolved significantly since 2016, when the Kremlin-linked Internet Research Agency bought targeted Facebook ads and organized staged rallies around divisive American issues. In the years since, Moscow has developed a more distributed and harder-to-trace infrastructure that combines state broadcasters, proxy media outlets, Telegram channels, bot networks, and now generative artificial intelligence. U.S. intelligence officials say Russia’s war in Ukraine has affected its capacity for foreign influence operations, diverting personnel and technical resources toward that conflict. Yet the war has also supplied the Kremlin with narrative tools, including claims that Western support for Ukraine is corrupt and that Washington is sacrificing American interests for a foreign war. This time, Russia is likely less interested in the midterm elections than in the next presidential race, according to intelligence assessments. Still, the current effort is being watched closely because even a small number of convincing AI-generated videos can produce outsized effects if they spread quickly through politically engaged communities.
The intelligence assessments also identify other foreign actors with potential motives to interfere in American elections. While Russian activity is the most visible, U.S. officials say Iran has shown interest in conducting influence operations as well, though its current footprint remains limited. Iranian operatives have historically engaged in cyber-enabled activities aimed at U.S. audiences, including hacking, phishing, and disinformation, but their operations have generally been smaller and more targeted than Moscow’s. China, according to U.S. officials, may be prepared to conduct campaigns aimed at individual congressional races or specific lawmakers whom Beijing regards as hostile to its interests. Chinese state-linked actors have previously been accused of using social media to shape narratives around human rights, trade, and Taiwan, and they could attempt to undermine candidates who have taken hard lines against China. The possibility of converging influence programs from multiple foreign adversaries has become a growing concern for U.S. national security agencies, which are monitoring the digital ecosystem for signs of covert activity. The assessments are based on intercepted communications, human intelligence, and analysis of openly visible online behavior. Officials acknowledge that measuring the efficacy of such campaigns is difficult; reach does not automatically produce persuasion, and many users may never encounter the material. Still, even one convincing deepfake spread through a popular account can generate millions of impressions and force mainstream media outlets to spend time debunking it. Responsibility for responding is fragmented across social media platforms, federal agencies, and state and local election officials, many of whom lack clear authority or sufficient resources to counter foreign interference. In this environment, intelligence officials expect disinformation to continue spreading in the weeks before the election. They caution that the current scale of Russian activity, while more limited than in previous cycles, could expand quickly if world events provide new opportunities or if Moscow decides to escalate its operations.
The broader significance of these fabricated celebrity videos lies in what they reveal about the changing mechanics of foreign interference. Earlier Russian influence campaigns relied heavily on armies of paid trolls writing inflammatory comments, setting up fake personas, and buying social media advertisements. Those activities were extensive, but they were often clumsy and were eventually exposed by journalists, researchers, and platform investigations. Today, generative AI allows a small number of operatives to produce a much larger volume of convincing content in less time and with fewer identifiable fingerprints. The videos supposedly showing famous actors denouncing particular candidates are cheap to make, difficult to attribute, and easy to move across different platforms. They exploit psychological shortcuts: a familiar face, a recognized newsroom logo, and an emotionally charged accusation are enough to make many viewers share the content without pausing to verify its origin. The episodes described in the NYT reporting illustrate how close technology has come to erasing the visual distinction between real and fabricated news. In one case, a video appeared to show Jamie Lee Curtis commenting on transgender politics; in another, a different celebrity was depicted making statements about a Senate race. None of the content was authentic, but the style, pacing, and graphics closely matched legitimate broadcast journalism. The effect is insidious: even after a specific video is exposed as fake, the public’s sense of what can be believed may remain shaken. Voters who learn that a viral clip was manufactured may become less trusting not only of social media but also of genuine news organizations, because the imitation was so seamless. American intelligence officials and outside researchers are therefore examining not just the content of the videos but also the infrastructure used to spread them: the timing of posts, the behavior of bot accounts, and the coordination between X and Bluesky. These networks are often resilient; when one account is suspended, new accounts are created to replace it. Some posts are removed after exposure, but by then they may have reached hundreds of thousands of users, and the debunking reports often reach a much smaller audience than the original fabrication.
As the United States enters the final weeks of the midterm election season, the discovery of a Russian-run deepfake campaign is a stark reminder that foreign adversaries continue to view American democracy as a target. The immediate challenge for platforms, law enforcement agencies, and election officials is to reduce the impact of the current wave without amplifying the very stories they are trying to counter. The longer-term challenge is to build a stronger immune system for the information environment: one that can identify manipulated media, expose coordinated networks, and equip citizens with the critical thinking skills needed to navigate a landscape in which seeing is no longer believing. U.S. officials say they are confident that the intelligence community and social media companies have improved their ability to detect foreign influence operations since 2016, but they concede that the race is far from won. The Russian campaign described by The New York Times may be smaller than previous efforts, but it is also more technically sophisticated and more deliberately targeted. Its emphasis on celebrity deepfakes and trusted-news branding suggests a tactical evolution: rather than trying to persuade broad numbers of voters, the attackers are attempting to contaminate the shared information space and force Americans to argue about what is real. That effect alone can be destabilizing. Unless voters, journalists, and elected officials remain vigilant, a foreign operation that began with a few fake celebrity videos could end by diminishing public confidence in the free press and the electoral process itself. The intelligence assessments do not suggest that Russia has abandoned its long-standing preference for chaos over any particular electoral outcome. But for American democracy, that distinction offers little comfort: chaos, once unleashed, is difficult to contain. In the coming weeks, officials expect more fabricated videos, more fake news articles, and more attempts to exploit the sharpest divisions in American politics. The only question is whether the public will recognize these materials for what they are: weapons in a quiet war being waged against the United States by adversaries operating far from its shores.
