2026-11-16 · The Hacker News · MEDIUM

AI Agents Target LiteLLM, Dify for API Keys

60-WORD SECURITY BRIEF

Attackers used autonomous AI agents to scan internet‑exposed AI platforms like LiteLLM, One API, and Dify for embedded API keys. Stolen keys authenticated to providers, allowing agents to run AI models that searched for more keys, creating a loop. Amazon’s threat team captured the activity with decoy servers and recommends revoking exposed keys, tightening storage, and enforcing stricter controls.