Microsoft Unveils MAI-Cyber-1-Flash AI Model for Autonomous Vulnerability Discovery and Patching in Minutes
Microsoft
Google/DeepMind
OpenAI
xAI
Anthropic
Microsoft introduced its first specialized cybersecurity AI model, MAI-Cyber-1-Flash, along with a new agentic platform Perception. The tools automate vulnerability discovery and patching in corporate environments. The model outperformed competitors in tests.
Microsoft has unveiled a specialized AI model named MAI-Cyber-1-Flash and a new agentic platform called Perception for cybersecurity. The model is designed to detect complex vulnerabilities in large codebases and works with the MDASH system for automated bug finding and patching. Perception deploys teams of agents to automate security workflows, using red, blue, and green teams to simulate attacks, identify errors, and fix them. Microsoft claims that MAI-Cyber-1-Flash outperformed models like Gemini, GPT 5.5 Cyber, GPT 5.6 Sol, and Mythos 5 in the Cyber Gym benchmark. The preview of MAI-Cyber-1-Flash and Perception will be available on November 3, competing with similar systems from Anthropic and OpenAI.
- Сокращения
- MDASH = Microsoft Defender Advanced Security Hunting? — система автоматического поиска и исправления ошибок Microsoft
Source: 3DNews —
original