Microsoft Unveils New AI Model for Cybersecurity Applications

Microsoft has launched a new artificial intelligence model, MAI-Cyber-1-Flash, specifically designed for cybersecurity applications. The announcement was made by Microsoft CEO Satya Nadella via a social media post on July 28.

The MAI-Cyber-1-Flash model is Microsoft's inaugural AI specifically focused on cybersecurity. It has been integrated into the company's MDASH system, which utilizes multiple AI agents. Microsoft states that the model can efficiently handle approximately 90 percent of cybersecurity tasks. For more complex challenges, the system is designed to invoke larger, more resource-intensive models.

In CyberGym benchmark tests, the combined MAI-Cyber-1-Flash and MDASH system achieved a success rate of 95.95 percent. This performance surpasses that of comparable models from competitors, including Anthropic's Claude Mythos 5, which scored 83.8 percent, and OpenAI's GPT-5.5 Cyber, with 85.6 percent.

Alongside the new model, Microsoft introduced Perception, a system comprising multiple agents that support various security workflows within MDASH. Perception is designed for continuous threat monitoring, vulnerability patching, and blocking new entry points for attacks. Microsoft plans to expand the roll-out of Perception across more security workflows in the future.

Microsoft indicated that the combination of MAI-Cyber-1-Flash and GPT-5.4 is projected to reduce the cost of the MDASH configuration by nearly 50 percent compared to previous setups. The company processes over 100 trillion security signals daily and employs reinforcement learning in its analysis of vulnerabilities, attacks, and countermeasures. The MAI-Cyber-1-Flash model has undergone rigorous evaluations, including assessments by Microsoft's internal AI red teams and third-party testers.