LLMjacking Attackers Rack Up $100K Daily in Cloud Fraud

LLMjacking Attackers Rack Up $100K Daily in Cloud Fraud
A joint report from Google's Mandiant and Google Threat Intelligence Group details a surge in LLMjacking, where attackers hijack enterprise cloud AI access and run up fraudulent charges exceeding $100,000 per day. Some victims faced cumulative bills of $1 million before access was revoked. Attackers gain entry via stolen credentials, exposed API keys scraped from GitHub, and misconfigured service accounts. Stolen access is resold through LLM Mixer operations that obscure usage origins. GTIG recommends spending caps, scoped API keys, and monitoring for single-token validation pings used to test stolen credentials.
Read the original article →