Google warns of rising theft of AI infrastructure access

Google researchers reported a sharp rise in "LLM-jacking," with stolen access to OpenAI, Anthropic and Google models resold on dark web markets at discounts of up to 97%. Attackers also breach enterprise cloud servers to run their own AI models, shifting computing costs onto victims.

• Stolen ChatGPT and Claude access sells at discounts up to 97%
• Premium subscriptions cost up to $200 per user per month
• Vendors offer "guaranteed access" with free credential replacement
• Anthropic logged Claude abuse attempts in over two dozen countries