Dark web markets offer access to AI models at steep discounts
Illegally obtained AI models and computing power are becoming a growing commodity in the cybercrime black market. Criminals aim to leverage stolen AI tools for extortion and espionage.
Illegally acquired AI models and computing power are rapidly emerging as highly sought-after commodities on the cybercrime black market. Criminals intend to leverage expensive large language models for extortion, cyber warfare, and espionage.
Google's threat intelligence team has observed a significant increase in what are termed "LLM hijackings" this year. These activities include the resale of stolen login credentials for public AI tools, as well as certain groups illicitly using others' computing power to run their own models for free.
According to research by Google's threat intelligence unit, dark web markets are selling access to AI models from companies like Anthropic, Google, and OpenAI at discounts of up to 97%. This contrasts with the highest subscription tiers for services like ChatGPT and Claude, which can cost up to $200 per user per month.
Some sellers are offering "guaranteed access" services, promising to provide new credentials at no extra charge if the initial account is banned. Furthermore, criminal groups and state-sponsored actors are infiltrating corporate cloud servers to install their own AI models, utilizing the victim's systems to run them—similar to how they exploit third-party devices for cryptocurrency mining.