|
||

Microsoft has introduced MAI-Cyber-1-Flash, its first AI model built specifically for cybersecurity, as part of MDASH, the company’s multi-model platform for identifying and remediating software vulnerabilities. The new configuration, which pairs the specialist model with GPT-5.4, reportedly achieved a 95.95% score on the CyberGym benchmark while reducing operating costs by about 50% compared with Microsoft’s previous MDASH setup.
Chief executive Satya Nadella described the launch as part of a broader effort to deliver “frontier-grade security at half the cost.” In a post on X, he said MAI-Cyber-1-Flash was designed from the ground up to identify challenging vulnerabilities in complex code bases and that combining specialized models, security data, AI agents and orchestration software enables Microsoft to improve the balance between cost and performance.
Benchmark claims: The company says the specialist model is intended to handle most routine security tasks, leaving only the most complex cases to GPT-5.4. However, the benchmark result applies to the combined MDASH system rather than MAI-Cyber-1-Flash alone, and the underlying cost calculations and evaluation criteria have not been fully disclosed. The reported score had also not appeared on CyberGym’s public leaderboard at the time of publication.
The launch forms part of Microsoft’s broader Project Perception initiative, an AI-driven cybersecurity platform built around specialized “red”, “blue” and “green” agents that simulate attacks, investigate threats and remediate vulnerabilities. Project Perception enters public preview on August 3, initially focusing on software vulnerability management before expanding into additional security workflows, reflecting Microsoft’s view that increasingly automated cyberattacks require equally autonomous defensive systems.
Sponsored byIPv4.Global
Sponsored byCSC
Sponsored byDNIB.com
Sponsored byWhoisXML API
Sponsored byRadix
Sponsored byVerisign
Sponsored byVerisign