Introducing MAI-Cyber-1-Flash inside MDASH
Microsoft introduces MAI-Cyber-1-Flash integrated into MDASH, a multi-agent system for vulnerability detection and remediation, offering 50% cost savings while outperforming rivals on CyberGym benchmarks.
Microsoft has unveiled MAI-Cyber-1-Flash, a new cybersecurity-focused model integrated into MDASH, its multi-agent vulnerability identification and remediation platform. The system is designed to address the escalating threat landscape by efficiently detecting vulnerabilities in complex codebases at a reduced cost compared to leading alternatives. MAI-Cyber-1-Flash is optimized to handle up to 90% of routine tasks, allowing MDASH to reserve larger, more expensive models for the remaining 10% of high-difficulty cases. The combined system achieves a 96% score on CyberGym, outperforming competitors like Mythos, Gemini, and GPT by 12 percentage points.
The announcement coincides with the launch of Perception, an agentic security system within MDASH that automates continuous monitoring, patching, and threat vector closure. Perception will soon expand its capabilities by incorporating MAI-Cyber-1-Flash for broader security workflows beyond vulnerability detection. Microsoft emphasizes that the system leverages decades of security expertise, trillions of daily signals, and a proprietary dataset of real exploits and remediations to deliver enterprise-grade protection.
MAI-Cyber-1-Flash is derived from the MAI-Thinking-1 lineage and was developed in-house using high-quality data, as detailed in a technical report. The model underwent rigorous security-first calibration, adversarial testing by Microsoft’s AI Red Team, and third-party validation to ensure reliability. MDASH provides enterprise controls including role-based access, tenant isolation, encryption, auditability, and sandboxed execution environments without internet access, aligning with Microsoft’s governance and security standards.
Microsoft frames cybersecurity as a live reinforcement learning loop, where defenders continuously investigate threats, remediate vulnerabilities, and refine defenses based on outcomes. The company processes over 100 trillion security signals daily from 1.6 million customers across identity, endpoint, cloud, and applications. This operational insight enables MAI-Cyber-1-Flash to improve continuously, reinforcing Microsoft’s commitment to building advanced, expert-level cyber defense models for long-term customer support.