# Memory poisoning: definition and example

URL: https://aiagentsecuritysolutions.com/glossary/memory-poisoning/
Updated: 2026-11-16

Memory poisoning is the planting of false or hostile instructions in the long-term memory an agent keeps between sessions. Because memory is read again later, one successful write can steer many future conversations. Defenses limit who can write memory and watch what it contains.

## Example

Zenity says researchers created malicious memories in AWS AgentCore that directed an agent's future conversations to an attacker-controlled destination.

## Sources

- [Zenity Labs reports AgentCorruption](https://zenity.io/press-release/zenity-labs-discloses-agentcorruption-a-chain-of-aws-agentcore-flaws) (zenity.io), read 2026-11-16

## Related pages

- [Goal hijacking](/glossary/goal-hijacking/)
- [Blast radius](/glossary/blast-radius/)
- [Agent trace](/glossary/agent-trace/)
- [AI Agent Security Risks, Mapped to Each Layer](/guides/ai-agent-security-risks/)
- [All glossary terms](/glossary/)
- [LLM gateway](/glossary/llm-gateway/)
- [Model routing](/glossary/model-routing/)
- [Guides](/guides/)
