Microsoft Says New Cybersecurity AI Model Helps MDASH Score 95.95% at Half the Cost

Microsoft has launched its first cybersecurity-specific model inside MDASH, its multi-model vulnerability identification and remediation harness. The company says MDASH, using MAI-Cyber-1-Flash and GPT-5.4, scored 95.95% on CyberGym. It also claims the configuration costs 50% less than its current best MDASH combination of GPT-5.4, GPT-5.4 mini, and GPT-5.3 Codex. Access is limited to approved

The article reports that Microsoft has launched MAI-Cyber-1-Flash, its first cybersecurity-specialized AI model, integrated into the MDASH multi-model vulnerability identification and remediation harness, achieving around 95.95–96% on the CyberGym benchmark while cutting MDASH configuration costs by about 50%.[1][3][7][9] Access to this configuration is limited to approved MDASH customers through an Azure AI Foundry private preview, and the model is only available inside MDASH rather than as a standalone public API.[1][7] From a RealGround perspective, this creates a concentrated dependency on a closed, multi-agent, multi-model security stack, raising AI supply chain risk around model provenance, configuration integrity, and update management. Organizations adopting MDASH and MAI-Cyber-1-Flash will need structured SBOM-style visibility and controls over how these agents and models are integrated, versioned, and governed to avoid hidden vulnerabilities or misconfigurations in the AI security tooling itself.