Microsoft Expands Project Perception to Monitor AI Agent Risks Continuously

Photo: Firstpost Tech

Microsoft is expanding Project Perception to continuously monitor AI agents, dynamically adjusting their permissions based on assessed risks. The project aims to contain potentially harmful actions by implementing security controls outside the AI models themselves, addressing concerns that frontier AI models cannot be fully trusted. This approach enhances safety by managing AI behavior in real time.

Tip: use ↑ / ↓ (or j / k) to move between stories.