Microsoft Debuts AI-Powered Cybersecurity Platform with OpenAI and Anthropic

Microsoft has unveiled Project Perception, a new AI-powered cybersecurity platform that integrates models from Microsoft, OpenAI, and Anthropic to detect vulnerabilities, analyze threats, and remediate issues. The platform's intelligent routing system determines which model is best suited for each task, allowing it to efficiently allocate resources and reduce costs.

Project Perception's performance was demonstrated on the CyberGym benchmark, where Microsoft's MDASH system scored 88.45% in May 2026, outperforming Anthropic's Mythos at 83.1% and OpenAI's GPT-5.5 at 81.8%. The platform is positioned as a lower-cost alternative to single-model systems, with Microsoft arguing that routing tasks to the most efficient model for each job reduces overall compute spend.

The move comes amidst recent developments in the cybersecurity space from Anthropic and OpenAI, including Anthropic's expansion of Project Glasswing and changes to OpenAI's Trusted Access for Cyber program. For enterprise buyers, competition is expected to drive better pricing, faster innovation, and stronger products.