Microsoft builds Project Perception to rival Anthropic's Mythos in security

answer: Microsoft is building a multi-model AI platform to compete with Anthropic's Mythos.

takeaways:

The Information reported that Microsoft is building Project Perception, an AI cybersecurity platform that uses several AI models to help organisations detect and repair software vulnerabilities, TechRepublic said on 17 July 2026.

Microsoft said the platform would use a model router to assign each task to the most suitable AI model, according to the report. Cheaper models would handle inventory, log parsing, codebase navigation and triage, while more capable models would reason through exploit paths and write fixes.

The platform draws on models from Microsoft, OpenAI and Anthropic, according to TechRepublic.

Project Perception is positioned against Anthropic's Claude Mythos 5, which the report said has strong vulnerability-hunting ability but is expensive and limited to vetted security partners. Mythos 5 is listed at $10 per million input tokens and $50 per million output tokens. Microsoft aims to match its analysis at lower cost.

The platform builds on Microsoft's earlier MDASH system, which scored 88.45% on the CyberGym benchmark in May 2026, compared with 83.1% for Mythos and 81.8% for GPT-5.5, CNBC reported.

Microsoft already sells the Defender, Sentinel, Entra and Security Copilot products. Project Perception would mark its first major security push since a leadership shake-up in February.

The platform was expected to be unveiled at a company event later in July, with a public preview to follow, according to the report.