Microsoft's 'Project Perception' Could Challenge Anthropic's Mythos in AI Security

What happened

Microsoft is developing an internal project codenamed Project Perception — a multi-model AI security platform combining models from Anthropic, OpenAI, and Microsoft. The project is led by Hayete Gallot, Microsoft's head of security.

Context and impact

Project Perception is directly targeting Anthropic Mythos 5, currently the leading AI tool for enterprise cybersecurity. Microsoft positions itself as a neutral AI model aggregator for security tasks.

Details

• Project: codename Project Perception
• Models: Anthropic + OpenAI + Microsoft (multi-vendor)
• Lead: Hayete Gallot (VP of Security, Microsoft)
• Goal: enterprise vulnerability finding cheaper than Anthropic Mythos 5
• Strategy: task routing to best-fit model by task type
• Status: in development, no confirmed launch date