Skip to content- Sources: primary, discussion
- Summary: Mandiant and Google Threat Intelligence report 2026 incidents where attackers used stolen IAM credentials, service account keys, GitHub PATs and leaked LLM API keys to enable model services and run inference, with charges above $100,000 in a day and up to $1 million in total. Stolen access is resold through LLM Mixers such as new-api based gateways and the jet-night.com service, and one actor escalated LiteLLM virtual keys to PROXY_ADMIN through CVE-2026-40217. Attackers validated keys with max_tokens=1 requests to avoid volume alerts.
- Why it matters: Teams holding model API keys or cloud credentials should alert on LLM service enablement and tiny validation requests, scope keys per model, set spend caps, and patch LiteLLM gateways.