Russia has launched new online influence operations targeting the U.S. election environment ahead of November’s midterm elections, including the use of AI-generated disinformation, The New York Times reported on Sept. 18, citing U.S. intelligence assessments, current and former officials, and disinformation researchers. Classified U.S. intelligence assessments in recent months concluded that the Kremlin had authorized another digital influence campaign aimed at spreading content on social media and amplifying divisions among Americans. U.S. officials have not found evidence that Russia is attempting to compromise voting equipment or directly interfere with the mechanics of voting. Instead, intelligence agencies are closely monitoring digital tools Moscow could use to shape public opinion and undermine confidence in the electoral process. The new operation represents the latest chapter in a pattern that has persisted since 2016, when Russian operatives first mounted a broad social media effort to exploit racial, political and social divides. It also arrives at a moment when generative artificial intelligence has made it possible for influence networks to create realistic but fake video, audio and images on a mass scale. The assessment is classified, a sign that the intelligence community is seeking both to understand the operation and to avoid revealing the methods by which it was detected. The immediate focus, officials say, is not on changing ballots but on changing minds. Moscow appears to understand that attacking the credibility of American elections can be as destabilizing as attacking the infrastructure itself. Because midterm elections often draw smaller and more partisan electorates, even a modest amount of disinformation can have an outsized impact in close Senate contests. The report also draws on interviews with former officials who described how the current campaign compares with earlier Russian efforts, providing a baseline for assessing whether Moscow is becoming more or less effective at interference.
Researchers in the United States and Europe recently detected renewed activity from a network previously linked to Russian influence operations. In two waves over the past week, the network distributed a series of apparently AI-generated videos on X and Bluesky containing false or unsubstantiated claims about Democratic candidates running for the U.S. Senate. Some of the videos were designed to resemble reports produced by major news organizations, including CNN, the BBC and The New York Times. The content included allegations of corruption and antisemitism and claims involving gay and transgender rights. The use of well-known news brands is a deliberate attempt to borrow the legitimacy of established journalism. A viewer who sees a fabricated CNN graphic or BBC-style report may share it without pausing to verify the source, especially if the content confirms pre-existing beliefs. The topics selected are equally deliberate: corruption attacks speak to concerns about political integrity, antisemitism claims are aimed at fracturing a Democratic coalition already divided over the Middle East, and gender-related stories are designed to ignite culture-war arguments that dominate social media. The choice of X and Bluesky is notable, as both platforms are heavily used by political elites, journalists and highly engaged voters. By publishing in two waves, the operation can observe which narratives are gaining traction and then adjust subsequent content accordingly. Threat-analysis company Graphika attributed the activity to a Kremlin-linked operation known as Operation Overload, also called Matryoshka, which has previously impersonated legitimate media outlets to spread fabricated stories. The dual name is fitting: like nested dolls, the operation embeds fake personas inside fake media platforms, hiding its Russian origin under layers of created evidence.
The campaign has targeted Democratic candidates in several competitive Senate races, including incumbent Georgia Sen. Jon Ossoff and candidates in Ohio, North Carolina, Maine, New Hampshire, Alaska, Michigan, Texas and Iowa, according to the Times. U.S. officials told the newspaper that the current Russian campaign appears less extensive and coordinated than some previous influence efforts. That assessment may reflect both improved detection capabilities and Russian adaptation. In 2016, Russia’s Internet Research Agency built sprawling networks of fake accounts and event pages that reached tens of millions of Americans. By 2018 and 2020, Moscow had shifted to smaller, more deniable operations focused on specific off-message narratives and vulnerable audiences. Operation Overload/Matryoshka is part of this evolution. Its use of AI-generated videos is particularly concerning because synthetic content can be produced at machine speed, personalized for different audiences, and rapidly altered to avoid detection. An AI-generated video costs a fraction of a traditional production and can depict events that never occurred, making it harder for fact-checkers to debunk every variation. The recent videos contained false or unsubstantiated claims about corruption, antisemitism, and gay and transgender rights. In Georgia, Ossoff is one of the most prominent targets because his seat is among the most contested in the country and because Georgia has become a focal point for broader arguments about election legitimacy. The inclusion of states such as Texas and Iowa, where Democrats are underdogs, suggests that Russia is not only trying to affect winners and losers but also to shape the national political environment and amplify the perception that the election system is rigged.
Russia has conducted online influence operations during every U.S. federal election cycle since 2016, according to previously declassified U.S. intelligence assessments. That continuous record has made Russia the reference point for measuring foreign interference. The current intelligence assessments focus primarily on attempts to influence perceptions and undermine confidence in American democracy rather than efforts to manipulate ballots or voting systems directly. This distinction is critical. It indicates that voting infrastructure is significantly harder to penetrate than it was a decade ago, and it suggests that Moscow has concluded that attacking public confidence is a cheaper and more effective strategy than hacking machines. If voters believe the electoral process is corrupt, the damage to democracy is done even if no vote is changed. The newest element in the current cycle is artificial intelligence. American officials and researchers view the rapid development of generative AI as one of the major emerging risks ahead of the November elections. AI tools allow foreign influence networks to produce fabricated videos, audio and other material more quickly and cheaply, potentially making it easier to run campaigns at scale. Voice cloning can simulate a candidate saying something they never said; generative video can place a politician in a compromising scenario; chatbots can automatically argue with voters in real time. The speed of AI-generated disinformation outstrips traditional fact-checking, so by the time a video is debunked it may have already been viewed and shared millions of times. Researchers warn that this lowers the barrier to entry for smaller foreign operations and even domestic political actors.
The New York Times report also notes that U.S. officials are monitoring possible influence operations by Iran and China, though their activities have so far been less visible in this cycle than Russian and Iranian efforts. China-linked activity has been comparatively restrained, according to officials and researchers cited by the newspaper, perhaps because Beijing prefers to promote longer-term narratives about American decline and hypocrisy rather than intervene directly in election contests. Iran, by contrast, has demonstrated a willingness to engage in aggressive cyber-enabled operations, including hack-and-leak campaigns designed to expose stolen information at politically damaging moments. The newspaper stressed that the scale, objectives and methods of foreign influence campaigns differ depending on the country involved. Russia’s overarching goal is often described as destabilization and the weakening of Western alliances; Iran’s objectives are linked more narrowly to retaliation for sanctions, support for Israel, and other bilateral grievances; China’s focus appears to be on reducing U.S. credibility on the world stage. These differences shape how officials assess risk. A Russian operation that fabricates compromising videos about Senate candidates is not the same as a Chinese operation that amplifies trade-war anxieties, but both can contribute to an information environment in which citizens distrust what they see and read. The absence of evidence that Russia is trying to manipulate voting systems is encouraging, but officials remain cautious. The intelligence picture is constantly changing, and new activity could emerge at any point before Election Day. The Times report is based on a snapshot of current findings, not a final judgment.
The discovery of the AI-generated videos has renewed debate about the responsibility of social media platforms and the adequacy of existing laws. X and Bluesky were the primary platforms used in the recent operation, but the same content can easily migrate to Facebook, Instagram, TikTok, Telegram and YouTube, where moderation policies vary widely. Many technology companies have scaled back trust and safety teams, and the rise of decentralized services makes enforcement more difficult. Some researchers and lawmakers argue that Washington should adopt a more aggressive deterrence strategy, including public attribution, sanctions against Russian intelligence entities, and cyber operations against the infrastructure used to spread disinformation. Others caution that overt retaliation against Moscow carries risks of escalation. In the absence of new legislation, the burden falls on social media companies, independent researchers, news organizations and voters themselves. Graphika’s attribution of the videos to Operation Overload/Matryoshka is an example of how open-source analysis can expose foreign interference, but detection is not the same as prevention. The public needs better tools for recognizing synthetic media, and journalists need to continue debunking fake content without amplifying it. The Times report is also a reminder that foreign disinformation succeeds only when domestic audiences share and amplify it. If U.S. politicians, influencers or ordinary users spread AI-generated content that aligns with their partisan interests, the operation will achieve its goal regardless of who originally created it. The ultimate defense, therefore, lies in the willingness of citizens and leaders to treat election information with skepticism and to preserve trust in democratic institutions. As the midterm elections approach, the challenge is to remain watchful without giving in to panic, and to protect the integrity of the electoral process without sacrificing the openness that defines a free society.