Dark Web AI Model Access Flooded with Discounts and Guaranteed Access Services
Cybercrime is evolving as hackers target AI accounts and computing power in a new wave of attacks known as 'LLM-jacking.' According to John Hultquist, principal analyst at Google Threat Intelligence Group, this type of attack has grown significantly this year. Dark web marketplaces are selling access to AI models from top companies like Anthropic, Google, and OpenAI at discounts of up to 97%.
Some sellers even offer 'guaranteed access' services, promising free replacement credentials if the initial account gets banned. This mechanism makes the illicit supply chain more resilient and difficult for AI companies to cut off. Hultquist noted that this mirrors earlier cryptojacking schemes, where hackers hijacked machines to mine cryptocurrency.
Another form of attack involves breaching enterprise cloud servers to deploy own AI models, forcing victims to foot the computing bill. Hultquist emphasized that the danger of this attack model lies in the economic imbalance, with attackers obtaining computing resources at costs far below normal market rates.