Memory Poisoning

Prevent untrusted content from becoming persistent agent memory.

Memory poisoning

CodeIntegrity gives teams policy controls for agent memory updates. Before memory is persisted, our policy engine checks the source, sensitivity, and workflow context, so untrusted content cannot become trusted memory.

Gate

Require policy approval before any agent memory update is persisted.

Inspect

Evaluate source, sensitivity, and workflow context for every memory write.

Block

Stop untrusted content from becoming trusted persistent memory.