Microsoft Announces Security Measures and Threat Warning for AI Agents

Microsoft has introduced new security measures, including Guardian Agents and Microsoft Defender updates, to protect companies against the security risks of artificial intelligence agents. The company stated that frontier models cannot be trusted and can behave like insider threats.

Microsoft has announced new security features to protect businesses against the growing threats posed by artificial intelligence agents. Introduced within this scope, Guardian Agents was developed to monitor the artificial intelligence agents used by companies. This feature, which is part of the Project Perception initiative announced in June, has been implemented at a time when trust in artificial intelligence providers is being questioned.

Hayete Gallot, Head of Microsoft Security Operations, stated that frontier models cannot be trusted because they are unpredictable and trained to be task-oriented rather than observing corporate norms. CEO Satya Nadella pointed out that the enemy of artificial intelligence may not be from the outside, but rather the agents that businesses have already deployed. Measures taken include Microsoft Defender updates to monitor agents and new features that dynamically limit agents' data access.