Privacy Leakage from Recurrent State in LLM Inference

Summary

Researchers study whether private information retained in the fixed-size recurrent state of latent-state LLMs can leak. Using partial-state probing and GPU power observation on released hybrid implementations, they find that a small state observation predicts private facts and response rules, that this information survives public dialogue, and that complete-read power can recover private facts under a fixed execution schedule. An unprivileged account on a partitioned GPU also inferred a user's age from whole-device power during public requests.

Related items

• MediumGHSA-68g8-gh8f-7rhg: pgEdge RAG Server discloses part of the configured provider credential through API error messagesSimilar attack · GitHub Advisory Database
• MediumGHSA-8835-xmqj-354h: pgEdge Go LLM Library discloses provider API key fragments in error messagesSimilar attack · GitHub Advisory Database
• MediumHackers stole AI credentials to get $740,000 in free compute; Railway picked up the billSimilar attack · CSO Online
• MediumOver 35,000 Grok-Built Apps Exposed Online, Some Leaking Confidential Data and PasswordsSimilar attack · The Hacker News
• MediumClaude Code’s ‘don’t ask again’ button has a secrets-spilling problemSimilar attack · CSO Online