Microsoft rolls out AI security tools, unveils Project Perception

On July 27, 2026, Microsoft introduced several AI-driven security initiatives aimed at combating AI-enabled threats. Key announcements included:

1. **Project Perception**: A new security system utilizing three types of agents (Red, Blue, Green) to identify vulnerabilities, investigate risks, and strengthen defenses.

2. **MAI-Cyber-1-Flash**: A generative AI model designed for cybersecurity and software vulnerability analysis, achieving a 95.95% success rate in CyberGym tests.

3. **Microsoft Security FORGE Lab**: A research lab led by Team Atlanta from DARPA's AI Cyber Challenge, focused on offensive security research and development.

4. **External Red Team Alliance (EXTRA)**: An initiative to fund AI safety research at 18 university labs globally, promoting diverse expert contributions to red teaming efforts.