AI-Powered Bot Armies Used by Iran, China, and Israeli Firms

Iran, China, and two private Israeli companies have recently employed artificial intelligence in an unprecedented manner to establish and manage autonomous influence campaigns on social media, according to a New York Times investigation. These AI agents reportedly created networks of fake accounts, selected topics for posts, disseminated political messages, and responded to users with minimal human oversight. American officials and security researchers told the newspaper that operators combined open-source Chinese AI models with autonomous "agents", software capable of operating other systems and completing tasks independently. Hundreds of these agents were deployed to create accounts on platforms like Facebook, Instagram, TikTok, and X, rapidly filling them with false political and current events content to sway public opinion and exacerbate divisions.

The core innovation highlighted is not the use of AI for generating deceptive text, images, or videos, which has been documented before, but the transfer of nearly the entire operational chain to the AI system. This marks one of the first instances where AI agents, based on Chinese models, managed campaigns from account creation to message coordination and online responses. In many cases, operators disabled the AI models' built-in safeguards against creating fake accounts or manipulating online discussions. The use of open-source models allows for operation on private servers and code modification, making detection and intervention by the developing company impossible.

While the campaigns described were still rudimentary, they attracted the attention of U.S. intelligence agencies, major tech companies, and security experts. Investigations revealed that the Chinese and Iranian campaigns were state-sponsored, while the two Israeli operations were linked to private commercial firms. The Iranian activity was deemed particularly concerning, with U.S. officials stating the network targeted an American audience, using AI-generated accounts posing as ordinary Americans in major cities. These accounts tagged journalists and politicians, spread popular memes and anti-Republican content, and attempted to infiltrate local political discourse, reaching nearly 80,000 followers during the first half of the year.

Fewer details were provided about the Chinese campaign, though intelligence and security sources indicated it was also state-sponsored and experimented with similar AI agent techniques. One of the Israeli-linked campaigns was reportedly run by IntelEye, a Tel Aviv-based company founded by former employees of the NSO Group. A U.S. security official briefed on the company's activities stated that its AI agents created thousands of accounts on social media platforms. These accounts responded to posts and comments related to Israeli elections, publishing content both supporting and opposing Prime Minister Benjamin Netanyahu. However, the investigation did not conclude that IntelEye specifically aimed to aid or harm Netanyahu, nor did it reveal who commissioned the activity. IntelEye confirmed using AI accounts and a Chinese model but denied running an influence campaign, stating their work was for defensive research and testing to identify vulnerabilities.

Meta confirmed identifying the activity, describing it as a coordinated attempt to create inauthentic accounts, and stated that most were removed by their automated systems. Meta's previous report also mentioned a separate campaign by an unnamed "commercial entity based in Israel" using a "fairly sophisticated system." The New York Times did not identify the second Israeli company involved. Both Meta and TikTok have reported increased AI use for account and content generation on their platforms, with Meta noting AI is now integrated into automated systems capable of producing and distributing content with minimal human intervention. X did not respond to inquiries.