The U.S. lacks a legal framework to govern AI-powered bot farms and coordinated political manipulation.
Sue Williamson and Ricardo Abbott were two ordinary Americans voicing their pro-Russian opinions on X, formerly known as Twitter, in 2024. Or were they? They were actually AI-generated personas, part of a Kremlin-backed bot farm disrupted by the Department of Justice in July 2024. That disruption was a rare win. The United States has no comprehensive legal framework to prevent the next one, and the threat has evolved faster than the law has moved to address it.
In September 2025, leaked documents from a Beijing-based technology firm called GoLaxy provided an unsettling window into just how far that evolution has gone. The documents described what the company called a “Smart Propaganda System”—a network of artificial intelligence (AI)-generated personas engineered to build psychological profiles of targets, adapt their messaging in real time, and saturate search results with fabricated content to warp the targets’ perceptions. One internal dossier revealed that the system had targeted 117 members of the U.S. Congress. The accounts did not just broadcast propaganda. They listened, learned, and persuaded.
GoLaxy was not alone: in June 2026, OpenAI disclosed that it had disrupted a Chinese influence operation that used its own tool—ChatGPT—to generate social media posts targeting U.S. public opinion on tariffs, AI policy, and data centers. Understanding why these operations are so difficult to detect and govern requires understanding how much the underlying technology has changed.
Traditional bot farms were labor-intensive—operators ran networks of fake accounts that mimicked human posting behavior, and their repetitive patterns made coordinated inauthentic behavior detectable by researchers and platforms. The new generation of AI-enabled influence operations has eroded many of those detection advantages.
A Frontiers in Artificial Intelligence study documents that AI-generated election disinformation is now indistinguishable from authentic human-written journalism in more than half of evaluated instances—meaning that the tools platforms and researchers built to flag synthetic content are increasingly inadequate. The same study found that bot-driven amplification accounts for roughly 25 percent of Twitter activity, and that AI-generated fake news websites have grown tenfold, with over 1,200 such sites identified by 2024 alone.
The persuasion dimension represents a further escalation. Research published in late 2025 found that AI chatbots shifted voters’ political views by a substantially larger margin than traditional political advertising. Bot farms are no longer simply amplifying messages that humans might otherwise ignore. They are actively constructing and updating in real time the information environment in which political judgments are made.
The existing U.S. regulatory architecture was not designed to address this threat. In 2024, the Federal Election Commission (FEC) declined to issue binding rules addressing AI-generated content in campaign communications, adopting only an interpretive rule that clarified existing misrepresentation prohibitions apply, regardless of the technology used. The FEC’s authority under the Federal Election Campaign Act applies specifically to candidates and their agents—it does not extend to political action committees, foreign state actors, or private firms operating AI persona networks from abroad.
Read full post
