Short · YouTube Shorts ·

Microsoft Beats the World’s Most Dangerous AI (Without Building One)

Microsoft Beats the World’s Most Dangerous AI (Without Building One)

Microsoft reported that MDASH topped the CyberGym security benchmark, outperforming Anthropic's Mythos Preview and OpenAI's GPT-5.5. The result was described as achieved without Microsoft building a new frontier model.

Read the full story at YouTube Shorts →