OpenAI has uncovered several international operations that misused its AI models for cyberattacks, political influence, and employment scams, according to a recent threat report. The activity spans countries from North Korea and Russia to Cambodia.
One scam detailed in the report involved using ChatGPT to draft job recruitment messages in multiple languages. Victims were promised more than five US dollars just for liking social media posts, a sum far above typical market rates, where 1,000 likes can cost less than ten dollars.
The fraudsters worked in a three-step process: they first attracted victims with unrealistic offers ("ping"), then built trust with fake testimonials and small payouts ("zing"), and finally demanded money for supposed entrance fees or cryptocurrency payments ("sting").
The operation, nicknamed "Wrong Number," was discovered accidentally when an OpenAI investigator received one of the initial messages via SMS. The message encouraged contact over WhatsApp, before forwarding the victim to a "mentor" on Telegram.
Russian propaganda targets Germany's federal election
OpenAI says it also shut down a Russian influence campaign that used ChatGPT to produce German-language content ahead of Germany's 2025 federal election. The campaign, dubbed "Operation Helgoland Bite," operated through a Telegram channel called "Nachhall von Helgoland" and an X account with over 27,000 followers.
The posts attacked the US and NATO and promoted the right-wing populist Alternative for Germany (AfD) party. ChatGPT was also used to research German opposition activists and bloggers, and to translate Russian-language propaganda into German. According to OpenAI, the campaign's overall reach was limited.
Chinese operations: From Taiwan to fake journalists
OpenAI identified several campaigns with ties to China. "Sneer Review" posted comments on geopolitical topics across TikTok and X, aiming to discredit critics such as Pakistani activist Mahrang Baloch.
Another operation, "VAGue Focus," used fake media accounts to collect information on Western targets. A third campaign, "Uncle Spam," flooded US political discussions with contradictory statements, such as on trade policy, and used AI-generated profile pictures of supposed military veterans.
OpenAI identified a total of ten international campaigns where threat actors turned its AI models toward fraudulent, political, or criminal ends.
