Software development was the first enterprise tech category to see a big impact from the arrival of AI agents, and thanks to developments like Project Glasswing and the Hugging Face incident, cybersecurity is quickly catching up. Despite its spotty record in cybersecurity circles in recent years, Microsoft hopes new security tools and models introduced Monday will draw attention and business away from the frontier labs.
The company unveiled its first AI model focused on security called MAI-Cyber-1-Flash, which was designed to work with Microsoft's MDASH security harness. It also introduced Project Perception, a collection of AI agents designed for offensive, defensive, and clean-up tasks that Microsoft placed at the center of what it is calling the new "Cyber Stack," a collection of sensor data, AI models, and agentic workflows.
