Generative AI accounts are being sold illegally on the dark web and Telegram at prices far below official subscription rates. With fraudulent offers and the risk of further credential theft, users are being warned to avoid these deals.
Cyber threat intelligence firm S2W said on Oct. 11 that an investigation using DRI (Deep Research Investigator), its autonomous AI agent designed for criminal investigations, found illegal sales of accounts for major generative AI services, including ChatGPT, Claude and Gemini, on the dark web and Telegram.
ChatGPT Plus accounts were offered for $2.20—about 11% of the official $20 monthly subscription fee. Claude Pro accounts were listed for $12-$15 a month, while 18 months of Gemini Pro access was being sold for around $10.
The sales spanned multiple channels. S2W identified related activity on more than 10 major dark web forums and more than five Telegram sales channels.
Alongside listings selling accounts and subscriptions, researchers found posts sharing accounts for free and seeking to buy accounts for specific AI services.
The trade also extends to API keys. These are digital credentials that allow software and services to access AI platforms such as ChatGPT and Claude automatically. If stolen, they can let someone use an AI service without the account holder's knowledge and run up charges in their name.
Growing AI adoption—and the costs that come with it—has helped create demand for these illicit offerings. As developers and businesses make greater use of AI APIs, their bills rise with the number of tokens processed, prompting some to seek cheaper access. Illegal sellers exploit that demand by advertising prices far below official rates.
Many of these deals, however, are tied to scams. Sellers reportedly offer accounts and API keys that have already been blocked or no longer work. Others pose as low-cost AI providers to steal users' own login credentials.
Even when buyers receive working access, continued use is far from assured. Major AI providers operate security systems that detect unusual access conditions or usage patterns, restrict account access and disable exposed API keys.
Experts say these strict security policies can sometimes trigger account blocks even within the same company when access conditions change. That makes sustained, reliable use of illegally purchased accounts effectively impossible.
“Cases have been reported in which attackers lure users with what appear to be low-cost AI services and persuade them to install software,” said Yang Jong-heon, head of S2W's integrated analysis division. “They then use that software to steal account credentials or API keys and resell them to other illegal AI service vendors, creating a vicious cycle.”
