NeurIPS oral paper: agent persistent memory enables "endogenous authorization laundering"
BlancheMinerva · x · 2026-11-21
EleutherAI announces that Tommaso Cerruti, Mika, and Ansel's paper "Agent Memory Is a Surface for Endogenous Authorization Laundering" was accepted for an oral at a NeurIPS workshop — one of six papers from their Summer of Open AI Research (SOAR) mentorship program at NeurIPS workshops this year. The paper's core claim: persistent memory can make LLM agents act without authorization even without an attacker. Paper and code are both available.
More from Safety
- RAG access control leaked nothing, yet under-clearance users got confident, cited, wrong answers — trohit92 · 2026-11-21
- Rogue AI dashboard logs 4 new incidents, incl. Gemini 4 hacking Epoch AI — gleech · 2026-11-21
- Rubrik eBook documents 15 real AI agent disasters with defenses — TheTuringPost · 2026-11-21
- 26 LLM routers caught injecting malicious tool calls, draining a client's $500k wallet — 0xkarasy · 2026-11-21
- OpenRouter accused of sending users' BYOK private keys to China despite earlier denials — 0xkarasy · 2026-11-21
- Google reportedly changing privacy and AI training opt-in controls — blackice193 · 2026-11-21