Skip to main content
Persistent agent memory stores unsanitized content that poisons future reasoning sessions

​
How to resolve

Scan all memory writes for malicious content before commit. Track provenance of memory entries. Use expiry and down-weighting for unverified memories.

​
Risk

Governance/Compliance This vulnerability falls under ASI06:2026 — Memory and Context Poisoning in the OWASP Top 10 for Agentic Applications. Persistent memory, RAG stores, or context summaries poisoned to alter future agent behavior. Security Persistent agent memory stores unsanitized content that poisons future reasoning sessions. If exploited, this can compromise the agent’s integrity, confidentiality, or availability, potentially affecting downstream systems and data.

​
Explanation

Persistent agent memory stores unsanitized content that poisons future reasoning sessions. This falls under ASI06 (Memory and Context Poisoning): Persistent memory, RAG stores, or context summaries poisoned to alter future agent behavior.

​
Specifications

Trigger
  • Agent scan detects persistent agent memory stores unsanitized content that poisons future reasoning sessions
The scanner looks for patterns such as:
  • memory=True with no content filter
  • store_memory() called on unvalidated output
  • append to memory without sanitization
Severity
  • HIGH
Applies to: Agent