Memory poisoning
Named by 8 kept items in this workspace.
7 items name it under techniques and 1 under classification.key entities.
Search every workspace for Memory poisoning. It matches the name as written, so it can find other things that share it.
The items that name it 8
agent/LLM long-term memory
Content written into memory once shapes later sessions, including other people's, named by OWASP as a distinct risk
False data enters an agent's short- or long-term memory and changes later decisions; the poisoned entry outlives the session that placed it
Cited by OWASP Top 10 for Agentic Applications 2026 as a critical threat category for agentic AI systems
Detected vulnerability category among the 17
Attack in which adversarially crafted content is stored in an agentic AI system's long-term persistent memory and subsequently influences future agent behavior, without modifying underlying model weights or system prompts; can suppress security alerts, facilitate privilege escalation, alter trust re
All techniques in this workspace.