> For clean Markdown of this page, append .md to its URL. For the complete documentation index, see https://www.superagent.sh/llms.txt.


Check whether untrusted knowledge base content can change a security-sensitive decision.

# AI agent knowledge base poisoning


[← Back to attack scenarios](https://www.superagent.sh/docs/products/red-teaming/attack-scenarios)

Agents often consult shared knowledge before they answer a question or take an action. This scenario checks how the agent handles untrusted material when a decision depends on what it retrieves.

It applies to agents that use retrieval for decisions about permissions or other protected workflows. A finding connects the untrusted information to the decision it influenced.

The app lists this scenario as **Poison shared knowledge used for security-sensitive decisions** (`AG-11`) in [Agent Red Teaming](https://www.superagent.sh/docs/products/red-teaming/agents).

## Standards and guidance

[OWASP Agentic Top 10 (2026), ASI06 Memory & Context Poisoning](https://genai.owasp.org/resource/owasp-top-10-for-agentic-applications-for-2026/)

---
Source: https://www.superagent.sh/docs/products/red-teaming/attack-scenarios/agent-knowledge-poisoning
Index: https://www.superagent.sh/llms.txt
