Attackers' AI agents stole API keys, then used them to pay for hunting more keys
krishnan · x · 2026-11-16
A 7-part thread breaks down Amazon's Nov 13 threat report (ARTEX AI agents): attackers' AI agents stole AI provider API keys, then used the stolen keys to pay for the compute of agents hunting more keys — an offloaded bill. Amazon's MadPot honeypots posed as LLM inference APIs; from Oct 3 to Nov 5, attacking agents sent prompts carrying their own campaign memory (system prompts, tool histories, findings). Targets: LLM gateways (LiteLLM, One API) and app builders (Dify, Langflow, Open WebUI). Entry points were mundane: default LiteLLM master keys, abandoned installs, unauthenticated endpoints, known CVEs, secrets in client-side JS. Keys public for 30+ days were graded low-value. Key takeaway: your agents also send full context to whatever inference endpoint they use — treat that route as a data boundary.
More from coding & agent
- User's $300 subscription generated $7,492 of usage — 25x value on Codex and Claude — theteknosaur · 2026-11-16
- Your agent's inference endpoint sees all its memory: treat it as a data boundary — krishnan · 2026-11-16
- Dev ships a gladiator management game built entirely with Claude Code — StarDrifterG · 2026-11-16
- Rust rewrite of mcbench-lite cuts first start from ~71s to under 1s with AI coding — adonis_singh · 2026-11-16
- Solo dev ships his son's game idea to iOS and Android for under $300 — morganinuk2 · 2026-11-16
- Enterprise AI coding fails on people, docs and process — not the codebase — _jaydeepkarale · 2026-11-16