Iran, China deploy autonomous AI agents to run influence campaigns

Iran, China, and private Israeli companies have begun deploying autonomous artificial intelligence agents to run social media influence campaigns with minimal human oversight, according to a New York Times investigation published Friday. The operations mark one of the first known instances in which AI managed nearly every step of an influence campaign — from creating fake accounts to posting content and interacting with real users.nytimes

AI Agents Replace Human Operators

Hundreds of AI agents, built on Chinese open-source models, were used to open fake accounts on Instagram Meta Platforms, Inc. ↘ 2.43%, Facebook, X, and TikTok and populate them with politically charged content, the Times reported. U.S. officials and cybersecurity researchers told the newspaper that the Iranian campaign was the most alarming: AI-generated accounts posed as ordinary Americans in major cities, tagged politicians and journalists, and circulated memes and anti-Republican Party content. Nearly 80,000 users followed accounts linked to the Iranian network during the first half of 2026.ynetnews+2

In several cases, safeguards designed to prevent AI systems from creating deceptive accounts had been disabled by the operators. Unlike earlier influence efforts that used AI merely to generate images or text, these campaigns relied on AI to autonomously create personas, select topics, publish material, and respond to other users.emirates247+1

Israeli Firm Admits to Controlling 1,000 Fake Accounts

The investigation also identified two private Israeli companies running similar operations. IntelEye, a Tel Aviv-based firm founded by former NSO Group employees, acknowledged purchasing 10,000 social media accounts and using the Chinese DeepSeek model to control roughly 1,000 of them, according to co-founder Maor Sellek. The accounts entered Israeli political discussions and posted content both supporting and opposing Prime Minister Benjamin Netanyahu.ynetnews

Sellek told the Times the activity was a "defensive research experiment" and not an influence campaign, though he acknowledged IntelEye should have informed the social media platforms. Meta called the activity "a coordinated attempt to create inauthentic accounts across several platforms" and said its automated systems had already removed the majority of the accounts.ynetnews

A Dangerous New Frontier

Security researchers stressed that while the campaigns remain relatively crude, the trajectory is clear. Kyle Crichton, a fellow at Georgetown University's Center for Security and Emerging Technology, told the Times that AI agents can "blend in and look more like a human user," posing risks beyond sheer scale. The use of open-source models is a particular concern because they can be downloaded, modified, and run privately, leaving no outside company able to intervene once safeguards are removed.emirates247+1

U.S. officials said the emerging capabilities have heightened concerns ahead of the November midterm elections.ynetnews