Microsoft MDASH AI Beats Anthropic in Cybersecurity Test

Microsoft’s new AI system, MDASH, has surpassed Anthropic’s Mythos on a leading cybersecurity benchmark. MDASH uses over 100 specialized AI agents working together to find real-world software vulnerabilities. What's interesting is, MDASH scored 88.45% on the CyberGym benchmark, outperforming Mythos Preview at 83.1% and GPT-5.5 at 81.8%. The system has already found 16 new vulnerabilities in Windows, including four critical flaws fixed this month. The bottom line is, while these AI tools can help find bugs, they also raise concerns about AI being used for offensive hacking, potentially leading to bigger software updates in the future.