NEWPractise supported commands in browser simulations, alongside the lesson.Try a browser practice exercise.Try a lab♥ Codelivly is 100% free, no paywall. If it helped you, chip in to keep it running.100% free. Help keep it so.Support us
Preview · 30 min
Master AI agent memory security. Learn how attackers exploit episodic memory poisoning and attention dilution, and defend with strict tenant isolation and sanitization.
A Large Language Model is stateless. Once an API call ends, it does not remember your last prompt. To make a stateful, autonomous agent, the system adds memory. Memory lets the agent learn from the past, keep user preferences and run workflows that last days.
For a defender, memory is also the best way for an attacker to persist. If long-term memory is poisoned, the payload does not need one clever prompt. It can sleep and trigger days or weeks later, when the agent retrieves it for a different, legitimate task.
The way agents recall state brings more risks: cross-tenant data leakage, attention dilution and indirect prompt injection through retrieved context.
Create a free account to unlock the rest of this task, complete tasks, and earn Rep.
Sign in