OpenAI has banned ChatGPT accounts used in an influence operation originating from Iran. This operation, dubbed "STORM-2035" by Microsoft, involved generating long-form web articles and short comments in English and Spanish.

The activity consisted of two main workstreams. One used OpenAI's models to create lengthy articles (700-900 words) for five websites, some posing as progressive and others as conservative, primarily focusing on the U.S. presidential election. The second workstream involved generating short comments for over a dozen fake personas on X and one on Instagram, discussing topics like Scottish independence, the U.S. election, the conflict in Gaza, and Venezuelan politics.

Meta confirmed the Instagram account's connection to an Iranian network they disrupted in December 2022. Despite the operators' attempts to research engagement tactics and even claim superiority over Russian and Chinese cyber operations, the content generated appears to have achieved minimal audience engagement. OpenAI's DALL·E 3 classifier also identified some generated images as not originating from their services.