How to meet it
OWASP AISVS v1.01
Memory, Embeddings & Vector Database Security · Embedding Sanitization & Validation
Validate what agents write to memory
Verify that no component reachable from agent execution can write agent outputs or tool outputs to trusted agent memory without explicit source validation.
OWASP AISVS v1.01, requirement 8.2.3
In plain English
Agents and tools should not be able to write into the agent's trusted memory without checking where the content came from.
What good looks like
- Writes to trusted memory go through a validation step.
- The source of each memory entry is recorded.
- Tool output is stored as untrusted unless validated.
How teams usually do it
- A memory service with separate trusted and untrusted stores.
Evidence an auditor would accept
- Memory architecture and write-path code.
Common mistakes
- Saving web page content straight into long-term memory.
Not sure which requirements apply to your system? Take the assessment.
Requirement text from the OWASP AI Security Verification Standard, licensed under CC BY-SA 4.0. This guidance is shared under the same licence.