Russia, Iran and China are using generative AI to push disinformation at American voters ahead of the US midterm elections, Bloomberg reported on October 14, 2026.
The midterms are scheduled for November 3, 2026. That leaves less than three weeks for platforms, AI companies and US officials to figure out how much of this content is out there, and who it is aimed at.
Who is doing what
According to the research findings, all three governments’ affiliated networks have turned to mainstream generative AI products, including OpenAI’s ChatGPT and Anthropic’s Claude.
Intelligence assessments from US agencies and independent analysts indicate the goal is not a single knockout blow. Instead, the operations aim to widen political divides, chip away at public trust in the election process and inflame tensions that already exist.
Russia is the most visible actor. A Russian-linked network known as Matryoshka, also tracked as Operation Overload and Storm-1679, has been spreading AI-manipulated videos since early September 2026.
Its output includes deepfakes of celebrities criticizing Democratic figures. It has also produced fabricated news clips styled to look like they came from established outlets such as CNN.
Iran has taken a different partisan angle. Iran-associated networks have used AI-generated material to push anti-Republican messaging, and some of them have posed as US activists.
Meta took down some of those Iranian networks earlier in 2026.
China is working at volume. Chinese operations tracked under the names Spamouflage and Green Cicada consist of thousands of artificial accounts, with AI used to manage those accounts and amplify their narratives.
The AI companies push back
OpenAI banned accounts tied to Russian and Iranian influence operations on October 8, 2026.
According to the research findings, OpenAI disrupted several Russian and Iranian campaigns that month. Those efforts involved inventing fake journalists and fake organizations, which were then used to spread misleading narratives.
Still, there are important limits on what is known. Investigators are still assessing the full scale of these operations and how directly they target the midterms specifically.
There is also a notable absence of evidence that any of these groups are trying to disrupt voting infrastructure directly. The threat, as currently understood, is to what voters believe, not to the machines that count their ballots.
How we got here
The Kremlin has run social media disinformation campaigns since 2016 to sow chaos in American elections, the New York Times reported on September 18, 2026.
What has changed is the toolkit. In 2016, influence operations mostly relied on amplifying messages through social media. A decade later, generative AI lets operators produce content cheaply and run campaigns at a scale that would have been hard to staff before.
What this means
OpenAI’s October bans show these companies can detect and remove bad actors. But detection after the fact means some content has already escaped into the wild.
Meta’s takedowns earlier in 2026 show enforcement is happening, yet the Chinese networks alone reportedly involve thousands of fake accounts.
According to the research findings, Washington has struggled to put comprehensive countermeasures in place against AI-enhanced tactics, and faces growing questions over whether it can keep pace with a threat that changes this quickly.
When a network fabricates clips designed to look like CNN, it borrows that outlet’s credibility.
What to watch next: whether more AI companies publish takedown reports, whether platforms announce additional network removals and whether US agencies release updated assessments before Election Day.
The other open question is whether investigators find any evidence of efforts aimed at election infrastructure itself. For now, the record points to a battle over perception rather than over polling places.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.