LATIDIA · Investigación

Lazare, veni foras: reprogramación de neuronas inactivas para agentes de IA de «puerta trasera»

Los investigadores de Microsoft han descubierto «Lazarus», una técnica de puerta trasera que puede reprogramar neuronas inactivas en un modelo de IA, lo que hace que un agente de IA, que de otro modo sería normal, ejecute comandos maliciosos cuando se activa un disparador oculto.

La noticia

Microsoft researchers have discovered “Lazarus,” a backdoor technique that can reprogram dormant neurons in an AI model, causing an otherwise normal AI agent to execute malicious commands when a hidden trigger appears. The threat is especially serious for AI agents with tool or terminal access. A triggered model could quietly download and run attacker code, steal secrets, or execute destructive commands while still passing standard safety evaluations. These backdoors can be inserted without retraining the model and are difficult to detect through normal testing, making model provenance and weight integrity critical security concerns. To protect customers, Microsoft Foundry verifies model provenance, ties models to trusted publishers, and scans the surrounding