Microsoft has unveiled MAI-Cyber-1-Flash, its first artificial intelligence model designed specifically for cybersecurity, alongside a new AI-powered framework called Project Perception. The company says the new technology can identify software vulnerabilities more accurately while reducing operating costs by nearly 50%.
The announcement marks Microsoft’s latest push to use AI not just for productivity, but also to combat increasingly sophisticated cyberattacks targeting businesses, governments, and critical infrastructure.
What Is MAI-Cyber-1-Flash?
MAI-Cyber-1-Flash is Microsoft’s first in-house AI model built exclusively for cybersecurity tasks. It is optimized to detect security flaws in software, analyze vulnerabilities, and assist in fixing them before attackers can exploit them.
The model operates within MDASH (Multi-Agent Vulnerability Identification and Remediation Harness), Microsoft’s AI-driven system that coordinates more than 100 specialized AI agents to discover, validate, and remediate software vulnerabilities.
Project Perception: AI Defending Against AI
Microsoft also introduced Project Perception, a next-generation security platform designed to help organizations respond to cyber threats at machine speed.
The platform continuously analyzes signals across an organization’s digital infrastructure, identifies potential risks, and enables AI agents to investigate and recommend or automate security responses. According to Microsoft, the first implementation focuses on software vulnerability management, with broader security applications planned in the future.
Higher Accuracy, Lower Cost
Microsoft claims the new cybersecurity model delivers impressive performance on industry benchmarks.
When integrated with MDASH, MAI-Cyber-1-Flash achieved a 95.95% (approximately 96%) score on the CyberGym benchmark, outperforming several competing AI systems used for vulnerability detection. The company also says the new setup reduces operational costs by about 50% compared with previous configurations by allowing the specialized model to handle most tasks while escalating only the most complex cases to larger AI models.
Built for Modern Cyber Threats
Microsoft says the rapid advancement of AI has changed the cybersecurity landscape, making attacks faster, cheaper, and more sophisticated.
Rather than relying on periodic software scans, the company believes organizations now need continuous AI-powered monitoring that can detect vulnerabilities in real time and respond before attackers exploit them. Project Perception is designed around this approach, combining AI reasoning, automation, and Microsoft’s vast security intelligence gathered from its global cloud services.
Public Preview Begins Soon
Microsoft announced that Project Perception and MAI-Cyber-1-Flash will enter public preview on August 3, allowing customers to test the new AI-powered cybersecurity capabilities. Initially focused on vulnerability management, Microsoft plans to expand the platform to additional security workflows over time.
A New Chapter in AI Security
With cyber threats becoming more complex and AI increasingly being used by attackers, Microsoft’s launch of MAI-Cyber-1-Flash signals a shift toward specialized AI models built for digital defense. By combining AI agents, automation, and real-time threat analysis, the company aims to help organizations detect vulnerabilities faster, strengthen their security posture, and reduce the cost of protecting critical systems.
