Person

Satya Nadella

1. Microsoft calls frontier AI agents 'insider threats', launches Guardian Agents to monitor them

   Microsoft on Monday released enterprise security features including Guardian Agents under Project Perception to monitor AI agents companies already deploy. CEO Satya Nadella recently warned that threats may come from agents firms themselves pay to use — like an insider — and said organisations should not rely only on model providers' safety assurances but build monitoring systems outside lab control. Project Perception already had red, blue and green team agents assuming external hackers; the new product responds to shifting risk views after Anthropic Mythos/Fable, OpenAI Daybreak and similar offensive–defensive models.

   Oct 20, 2026 Afternoon