poisoned memory

Poisoned memory is the security problem that arrives the moment an AI agent stops forgetting. New research published in September 2026 argues that once an agent writes what it learns into a store it will read back later, an attacker no longer needs to win the conversation in front of them. They only need to […]