Russian Disinformation Campaign Targets US Midterms with Celebrity Deepfakes
The Emergence of a New Influence Operation
A sophisticated Russian disinformation operation has shifted its focus to the upcoming US midterm elections, deploying a coordinated series of artificial intelligence-generated deepfake videos that depict prominent Hollywood celebrities attacking Democratic Party officials and candidates. The videos, which began circulating across multiple social media platforms late last week, feature fabricated audio and manipulated imagery designed to appear as authentic celebrity endorsements of anti-Democratic messaging. The content carries hashtags such as “#AllDemocratsAreCriminals” and has been branded with CNN’s logo to lend false legitimacy to the material, creating a dangerous fusion of celebrity influence and fabricated media that researchers warn could significantly impact voter perceptions in the critical November elections.
The deepfake videos employ sophisticated voice-cloning technology to create convincing but entirely fabricated statements from well-known entertainment figures. One particularly troubling video shows “How I Met Your Mother” star Alyson Hannigan appearing to make inflammatory statements about children being endangered by Democratic policies, while another depicts actress Emma Caulfield making bizarre claims about Democrats drinking blood. A third video features “Sex and the City” star Sarah Jessica Parker apparently calling Ukrainian President Volodymyr Zelensky and Democrats “Nazis.” These fabricated statements are designed to exploit existing political divisions and amplify extreme rhetoric through the trusted voices of beloved entertainment figures, making the content particularly dangerous for unsuspecting viewers who may not immediately recognize the manipulation.
Technical Methods and Exploitation of Trust
The creation of these deceptive videos relied on the theft and manipulation of legitimate footage sourced from various corners of the internet, raising serious concerns about the vulnerability of publicly available content. One particularly sophisticated fake layered fabricated audio over a 2023 Gucci advertising campaign featuring actress Julia Roberts, demonstrating the technical sophistication of the operation. More troubling still, other videos were constructed using authentic recordings uploaded to Cameo, the popular service that allows celebrities to sell personalized video messages to fans. This exploitation of legitimate commercial platforms represents a new frontier in disinformation tactics, as the operators have found ways to weaponize the very services that celebrities use to connect with their audiences authentically.
The victims of this technological exploitation have expressed profound distress over the manipulation of their images and voices. Caulfield, known for her role in “Buffy the Vampire Slayer,” described feeling violated by the fabricated video, emphasizing that the artificial creation bears no resemblance to her actual beliefs or personality. Similarly, actor Ed Begley Jr., whose Cameo recording was altered to depict him making derogatory statements about Democratic politicians, expressed complete shock at discovering the manipulation. In a statement to AFP, Begley emphasized that he wholeheartedly disagrees with everything the fabricated video claims. “Runaway Train” star Eric Roberts, whose Cameo footage was similarly distorted to attack Democratic positions on transgender issues, told reporters that the statements “are not my words and do not represent my beliefs,” while expressing deep concern about the unauthorized use of his likeness to deceive the public.
Attribution and Pattern Recognition
Disinformation researchers have identified the operation as bearing the distinctive hallmarks of Matryoshka, also known as Operation Overload, a known Kremlin-linked campaign that specializes in creating and distributing doctored videos impersonating legitimate news organizations. The campaign, which has a documented history of targeting elections across multiple Western democracies, uses sophisticated bot networks to amplify its content across social media platforms, creating the appearance of organic grassroots support for the fabricated narratives. Darren Linvill, an expert on Russian influence campaigns at Clemson University, characterized this as “the first salvo in Russia’s targeting” of the November vote, noting that Russian operatives have systematically targeted every US election with social media manipulation campaigns since 2016.
The online research collective Antibot4navalny first identified the connection to Matryoshka based on the campaigns’ distinctive patterns of distribution and amplification across social media platforms. These patterns include the characteristic timing of posts, the use of coordinated bot networks to boost engagement metrics, and the strategic deployment of hashtags designed to maximize visibility. The use of Cameo videos represents a proven tactic for the group, which has previously employed similar methods in elections across Europe, including regional and local contests in Germany, Hungary, and Armenia. French prosecutors have also launched investigations into suspected Russian deepfake operations targeting centrist candidates in their upcoming presidential elections, indicating that this threat extends well beyond American shores.
Strategic Implications for the Midterms
The deepfake campaign signals a calculated shift in Russian influence operations toward the issues and races that will determine control of the US Congress in the November elections. The fabricated videos appear designed to exploit existing cultural and political fault lines, particularly around issues such as transgender rights, child welfare, and foreign policy. The inclusion of references to Ukraine and Zelensky suggests that the operation seeks to capitalize on divisions over American support for Ukraine’s defense against Russian aggression, a topic that has become increasingly polarizing in American politics. By associating Democratic candidates with “Nazi” imagery and extreme positions, the operation aims to drive a wedge between moderate voters and the Democratic Party while energizing the most extreme elements of the conservative base.
The operation has expanded beyond celebrity endorsements to include fabricated news reports targeting specific Democratic candidates in key races. Researchers have identified additional fake CNN reports smearing Georgia Senator Jon Ossoff and Texas Senate hopeful James Talarico, suggesting that the operation is methodically targeting candidates in competitive races that could determine which party controls the Senate. The group has also been credited with producing inauthentic video reports designed to discredit legitimate fact-checking efforts, including fabricating content that claims journalist Brandy Zadrozny has retracted her coverage of the disinformation campaign. This reflexive approach, where the operation seeks to undermine reporting on its own activities, represents an increasingly sophisticated understanding of the information ecosystem and the importance of establishing narrative control.
Response and Enforcement Challenges
The dissemination of this fabricated content has triggered response efforts from multiple stakeholders, though questions remain about their effectiveness. CNN has explicitly disavowed the videos, with a spokesperson emphasizing that the fabricated content “are not CNN content and have no affiliation with CNN” and that the network objects to the unauthorized use of its branding. Some social media platforms, including X (formerly Twitter), Bluesky, and TikTok, have removed certain posts containing the deepfake content, demonstrating that platform enforcement mechanisms can work when properly deployed. However, the viral nature of the content and the automated systems used to distribute it make complete removal challenging, forcing researchers and platforms to engage in an ongoing game of whack-a-mole as new variations appear.
The affected celebrities have also begun speaking out publicly to distance themselves from the fabricated statements, though they face the challenge of correcting misinformation that may be much more widely shared than their rebuttals. The sight of multiple celebrities simultaneously issuing denials could itself become part of the story, potentially amplifying the original disinformation by driving curiosity about the fabricated content. This challenge reflects a fundamental asymmetry in the information environment: fabricated content can be created and distributed quickly and cheaply, while verification efforts are necessarily slower and more expensive. The persistence of the operation, with new fabricated videos appearing even after initial content was removed, suggests that the perpetrators are adapting their tactics in real-time to evade detection and enforcement efforts.
A Global Pattern of Election Interference
The celebrity deepfake operation represents the latest manifestation of a broader pattern of Russian election interference that has evolved significantly since the 2016 presidential election. While early operations relied primarily on social media bots and human trolls to spread divisive content, current operations leverage sophisticated artificial intelligence technologies to create persuasive fabricated media that is virtually indistinguishable from authentic content. The evolution of these tactics reflects a learning process on the part of Russian operatives, who have refined their approaches based on the successes and failures of previous campaigns. The targeting of both US and European elections suggests that election interference has become an institutionalized component of Russian foreign policy rather than an improvised tactic.
The timing of the operation, beginning months before the November elections and showing signs of increasing intensity, suggests that the current wave of deepfakes represents only the initial phase of a comprehensive campaign. Researchers anticipate additional fabricated content emerging as the election grows closer, potentially targeting different states, candidates, and issues as the political landscape evolves. The low production costs associated with creating deepfake content, particularly the exploitation of Cameo videos that can be scraped and manipulated cheaply, suggest that the operation can produce large volumes of content without substantial financial investment. This cost asymmetry poses fundamental challenges for democratic societies attempting to protect the integrity of their elections while preserving the benefits of open internet communication and free expression.
The ongoing threat of Russian election interference through sophisticated deepfake technology requires a coordinated response from governments, technology companies, media organizations, and the public. Understanding the mechanisms of these operations, including their reliance on stolen content, branded replication of legitimate news organizations, and coordinated amplification through bot networks, provides essential tools for detecting and debunking fabricated media. However, the resilience and adaptability of the operations suggest that complete prevention is unlikely, requiring citizens to become more discerning consumers of information and news organizations to maintain vigilance against the weaponization of their brands and methods. As the midterm elections approach, the emergence of these matryoshka-style operations serves as a stark reminder of the ongoing threats to democratic processes in the digital age.