Researchers expose hidden backdoor that hijacks dormant AI neurons
What is Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents?
Microsoft researchers have demonstrated a technique called Lazarus that can reprogram inactive neurons inside an AI model. When a secret trigger appears, the model silently follows attacker-specified commands.
Who it is for
AI developers and security teams need to know this vulnerability so they can audit models and defend against hidden backdoors.
The story so far
Early Signal first spotted Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents on November 11, 2026. Early Signal files it under Agents.