Microsoft Prepares Project Perception to Compete With Anthropic Mythos

Microsoft is reportedly developing an AI-powered cybersecurity platform called Project Perception, according to a report from The Information. The tool could help organizations detect and repair software vulnerabilities using multiple artificial intelligence models.

Project Perception could launch later this month, although Microsoft has not officially confirmed the reported release window or the platform’s full capabilities.

Project Perception Could Use Multiple AI Models

Project Perception will reportedly combine AI models developed by Microsoft, OpenAI, and Anthropic.

The platform could use a model router to select the most suitable model for each cybersecurity task. One model might perform better at reviewing source code, while another could identify unusual behavior or suggest fixes for discovered vulnerabilities.

Using several models may give Project Perception broader vulnerability coverage than a system built around a single model. Microsoft could also route simpler tasks to less expensive models, helping reduce the platform’s operating costs.

Microsoft Reportedly Wants to Compete With Anthropic Mythos

Microsoft reportedly wants Project Perception to offer capabilities similar to Anthropic’s Mythos model.

Mythos is said to provide advanced vulnerability-hunting capabilities, but it can also require substantial computing resources and remain expensive to operate. Microsoft reportedly hopes to deliver comparable security analysis more efficiently through its multi-model architecture.

Microsoft has not finalized pricing for Project Perception. The company could offer the platform as a standalone product, include it in existing security subscriptions, or restrict access to selected enterprise customers.

Project Perception May Support Microsoft’s AI Security Strategy

The project is reportedly being led by Hayete Gallot, who became Microsoft’s security chief in February 2026.

Gallot has reportedly been reorganizing Microsoft’s security division around products that use artificial intelligence to identify threats, investigate incidents, and repair software weaknesses.

Microsoft is also said to be reducing investment in some older security offerings as it shifts resources toward AI-powered products. Business-focused AI security services could help the company recover part of the substantial amount it has spent on infrastructure, model development, and data centers.

Microsoft Faces Growing AI Cybersecurity Competition

Microsoft already operates one of the world’s largest cybersecurity businesses, offering products such as Microsoft Defender, Sentinel, Entra, and Security Copilot.

Project Perception could strengthen its position as Anthropic, OpenAI, and other technology companies develop AI agents capable of reviewing code and identifying security flaws.

More competition could lower the cost of advanced vulnerability analysis for businesses. It may also encourage providers to make capable security models available to more organizations instead of limiting them to a small number of customers.

Access Could Initially Be Restricted

Microsoft may initially limit access to Project Perception because vulnerability-hunting AI systems can create significant security risks.

A capable model could discover exploitable software flaws before developers release patches. Attackers could misuse the same technology to locate weaknesses in widely deployed applications, cloud services, or infrastructure.

Microsoft could therefore restrict the platform to vetted companies, government organizations, security researchers, or customers that meet specific compliance requirements.

The company may also introduce monitoring systems and usage limits designed to prevent customers from using Project Perception against unauthorized targets.

Possible Connection to MDASH Remains Unclear

Microsoft has also expanded its AI security work through MDASH, although it remains unclear whether Project Perception connects to that initiative.

The two systems could share technology or operate as parts of the same broader security strategy. However, Microsoft has not confirmed any relationship between them.

The reported project comes as Microsoft continues addressing vulnerabilities across its software portfolio. During the July 2026 Patch Tuesday release, the company fixed 570 security vulnerabilities affecting Windows and other Microsoft products.

Microsoft has not publicly announced Project Perception, so its final name, availability, supported models, and pricing could change before launch.

Via Neowin