METR API key stolen and abused for three weeks, racking up ~$600,000 in credits
Kyrannio · x · 2026-09-14
A real-world AI security incident: a fail-open bug disabled Google authentication on a public agent dashboard, letting attackers prompt the agent into revealing its API key and add SSH persistence.
- The key belonged to AI safety eval org METR and was abused for roughly three weeks
- Attackers consumed about $600,000 in credits
- The case underscores fail-open auth risks and prompt injection exposure in deployed agent systems
Related event: METR Breached: Attackers Burn ~$600K in API Credits Over Three Weeks(9 posts)→
More from Safety
- Frog and Toad meme nails the HSM paradox: keys stay safe, but compromised servers can still sign — rickasaurus · 2026-10-31
- Heads Of Security: paper pinpoints and rewires attention heads behind prompt injections — jonasgeiping · 2026-10-31
- House Democrats resolution: no bailouts for failing AI companies — your data goes to bankruptcy court — alex_verem · 2026-10-31
- Amazon Echo kills local voice storage: everything goes to the cloud from March 28 — alex_verem · 2026-10-31
- DOJ's first crackdown on AI-assisted "ghost students" seizes $750K+ in crypto — Ars Technica AI · 2026-10-31
- Researcher Finds 600+ Bugs in LLM Kernel Code; Claude Blocks Safety Analysis — pmddomingos · 2026-10-31