Microsoft introduces its own AI model for cybersecurity

Microsoft is expanding its in-house MAI model family with an AI model specializing in cybersecurity. The company announced the expansion of its own AI models, introducing MAI-Cyber-1-Flash, a model specifically designed to detect security vulnerabilities in complex codebases. According to Microsoft's announcement, MAI-Cyber-1-Flash is a compact, highly code-focused security model that originates from the MAI-Thinking-1 line and was developed entirely internally on its own training data. The model is based on a Sparse Mixture-of-Experts Transformer architecture with a total of 137 billion parameters, of which only 5 billion are active simultaneously, and has a context window of 256,000 tokens. Microsoft is integrating the model into MDASH, a multi-agent system with over 100 specialized agents, where MAI-Cyber-1-Flash handles around 90 percent of tasks and escalates the most difficult 10 percent to OpenAI's GPT-5.4. This combination achieves a hit rate of 96 percent on the CyberGym benchmark and reduces costs by around 50 percent compared to the previous MDASH configuration.