WebTechKitchen; Your Web Technology Kitchen, contact us to create, or maintain your websites and other digital properties.

We build websites, manage SEO, security, and all your digital concerns

Microsoft Built a Cheaper Mythos and Called It Project Perception

AI Tech News

Microsoft Built a Cheaper Mythos and Called It Project Perception

· By aiuser

Anthropic's Claude Mythos Preview walked so a bargain-bin AI security scanner could sprint. Microsoft is reportedly getting ready to launch its own answer to Anthropic's AI-powered bug hunter — and the whole pitch is basically "same job, smaller invoice."

One Router, Three Vendors' Models, Zero Loyalty

Project Perception is an AI-powered platform that scans enterprise codebases for exploitable vulnerabilities, and it's expected to launch before the end of July. Rather than betting on one frontier model, it routes tasks through a model-selection layer pulling from Microsoft, OpenAI, and Anthropic's own models — reserving expensive frontier calls for the steps that actually need them and handing high-volume scanning passes to cheaper, distilled models.

That architecture is a direct shot at Mythos's pricing, which runs $10 per million input tokens and $50 per million output tokens — the highest rate of any generally available model on the market. Microsoft hasn't announced official pricing for Project Perception yet, and it's unclear whether it'll ship standalone, bundled into existing security subscriptions, or gated to select enterprise customers first.

The Real Fight Isn't Accuracy, It's the Bill

This is a familiar pattern: whoever ships the flashy frontier capability first eventually gets undercut by whoever's willing to build the cheaper, multi-vendor version that's "good enough." Continuous, always-on vulnerability scanning across an entire enterprise codebase gets expensive fast at frontier-model rates, and Microsoft is betting that cost, not marginal accuracy, is what actually decides enterprise security budgets.

It's also a quietly funny moment: Microsoft's AI security tool will lean on Anthropic's own models as part of its routing stack while competing directly against Anthropic's flagship security product. Coopetition, but make it ironic.

If Project Perception lands anywhere near its promised price point, expect a lot of CISOs to suddenly discover they never really needed the expensive option.

Source: Tech Times