Microsoft AI releases cybersecurity model MAI-Cyber-1-Flash, improving vulnerability detection performance and reducing costs by 50%
Microsoft
Microsoft AI has launched its first dedicated cybersecurity model, MAI-Cyber-1-Flash, and integrated it into the multi-agent vulnerability identification and remediation platform MDASH. The combination achieved a 95.95% score on the CyberGym vulnerability detection benchmark, surpassing Mythos, Gemini, and GPT series models, while reducing costs by approximately 50% compared to the existing MDASH solution.
On July 27 local time, Microsoft AI released MAI-Cyber-1-Flash, its first cybersecurity-specific model, and integrated it into the multi-agent vulnerability identification and remediation platform MDASH. In CyberGym vulnerability detection benchmark tests, the combination achieved a 95.95% score, outperforming Mythos, Gemini, and GPT series models while reducing costs by about 50% compared to the existing MDASH solution. Additionally, Microsoft introduced the Perception security system for agents, which continuously monitors, patches vulnerabilities, and responds to new threats.
Source: 36Kr β
original