memory-poisoning

2 live · 3 on record · last incident 3 days ago

Dec ’25records per month, by first reportNov ’26

Live2

live
On Private Knowledge Retrieval and Private Inference for Multi-Agent Systems
Research · since 11 Nov 2026
Researchers explore memory poisoning and security vulnerabilities within multi-agent systems built on large language models. The study details attack vectors targeting agent knowledge and retrieval components.
AI
live
LLM Agent Memory Managers Accept Changes, Reject Attacks: Robust Memory Management for LLM Agents
Research · since 2 Nov 2026
Researchers explore memory poisoning and outdated memory vulnerabilities in long-term memory management systems for LLM agents. The study demonstrates how planted passages and injected instructions can compromise agent memory integrity.
AI