OpenAI bans Israeli firm STOIC for running a sprawling, zero-engagement influence op

OpenAI has disrupted a previously unreported influence operation originating from Israel, dubbed “Zero Zeno,” that generated AI-powered propaganda across multiple social media platforms and websites. The network was operated by STOIC, a political campaign management firm based in Israel. The name “Zero Zeno” is a nod to both the stoic school of philosophy and the campaign’s abysmal performance—it attracted virtually no authentic engagement.

The operation was a chaotic thematic mess. It used OpenAI’s models to spin up fake personas, generate social media comments, and pen web articles targeting audiences in Canada, the U.S., and Israel with anti-Hamas, anti-Qatar, and pro-Israel narratives. In a bizarre pivot in early May, accounts suddenly began criticizing India’s ruling BJP party and praising the opposition Congress. The network even appeared to be gearing up for a campaign in Ghana before it was banned. The content was seeded across Facebook, Instagram, and X, as well as four websites masquerading as activist groups, including uc4canada.com and ufnews.io.

The operational security wasn’t exactly sophisticated. Multiple accounts used the same GAN-generated profile pictures, and some fake accounts even replied to their own network’s posts—a clumsy attempt to simulate grassroots support. At one point, a comment about Qatar “buying up” America was nonsensically posted in reply to a social media post about Donald Trump’s taste in music. Researchers from the Atlantic Council’s Digital Forensic Research Lab linked the network to content criticizing the UN relief agency UNRWA and attacking pro-Palestinian protesters on U.S. campuses as promoters of antisemitism.

Here’s how I see it: this operation is less a demonstration of AI’s propaganda power and more a case study in how not to run a covert influence campaign. The complete failure to generate organic reach is almost comical, but it masks a more serious reality. The low-effort spray-and-pray approach, flitting between geopolitically charged topics from Gaza to Canadian Islamophobia to Indian elections, suggests the barrier to entry for state-linked IO is now just the cost of an API key. The fact that STOIC was a commercial firm, not a government agency directly, also blurs the already murky line between mercenary digital services and state-directed information warfare. While OpenAI’s disruption was swift in this case, the next operation might be run by someone who actually understands how to build an audience first.