Microsoft Project Perception: AI agents monitored by Guardian agents, risk of cascading failures

Microsoft's Project Perception introduces Guardian Agents to monitor and contain anomalous AI agents, aiming to unify security across agentic AI workflows. Experts caution that using AI to police AI may amplify unpredictable behavior, creating cascading risks. The initiative signals a major enterprise security shift but its efficacy remains unproven.

AI disclosure: Model-written summary of the source, not a verified fact. Check the sources. How Mosaic works