Microsoft is working on AI-powered cyber defence tool to take on Anthropic's Mythos: Report

The race to build the next generation of AI-powered cybersecurity tools is gathering pace, and Microsoft appears ready to make its next move.

According to a report by The Information, the technology giant is preparing to launch Project Perception, a new security product designed to help organisations identify and fix software vulnerabilities with the assistance of artificial intelligence. The product could debut as early as this month and is expected to combine AI models from Microsoft, OpenAI and Anthropic, allowing it to tackle different security tasks while keeping operating costs under control.

If launched as reported, Project Perception would mark Microsoft's latest effort to strengthen its enterprise security portfolio at a time when businesses are increasingly investing in AI-driven cyber defences to counter a rapidly evolving threat landscape.

A multi-model approach to finding software flaws

Unlike security tools that rely on a single AI model, Project Perception is expected to use what is known as a model router. According to The Information, this system would determine which AI model is best suited to a particular task before assigning the workload accordingly.

More from Tech

The approach would allow Microsoft's own AI models, alongside those from OpenAI and Anthropic, to work together rather than independently. Depending on the complexity of a software vulnerability, the platform could switch between models to analyse code, identify security weaknesses and automatically generate fixes.

The strategy is also intended to reduce costs. Anthropic's cybersecurity-focused AI model, Mythos, has gained attention for its advanced bug-hunting capabilities but is considered expensive to deploy at scale. By routing requests across multiple models instead of relying solely on one, Microsoft reportedly hopes to offer customers similar capabilities at a lower price point.

Pricing for Project Perception has not yet been finalised, according to the report.

The timing reflects a broader shift within enterprise cybersecurity, where AI is increasingly being used to automate tasks that previously depended on large teams of human analysts. Modern organisations face an ever-growing number of cyber threats, while attackers themselves are also beginning to use AI to discover software vulnerabilities more quickly.

Part of Microsoft's broader AI security push

Project Perception is understood to be one of the first major initiatives under Hayete Gallot, Microsoft's new head of security, who assumed leadership of the division earlier this year.

Since taking charge in February, Gallot has reportedly reorganised Microsoft's security business to place greater emphasis on AI-powered products while reducing investment in older offerings. The changes are aimed at reinforcing Microsoft's leadership in enterprise security as competition intensifies.

Although Microsoft remains the world's largest provider of enterprise security software, newer AI-focused rivals have rapidly attracted attention. Anthropic, in particular, has emerged as a significant player after demonstrating specialised AI systems capable of identifying software bugs and other cybersecurity risks.

That shift has fuelled growing demand from businesses looking to strengthen their cyber defences against increasingly sophisticated attacks, especially those that are themselves being enhanced by artificial intelligence. Security executives have told The Information that organisations are investing heavily in AI tools capable of continuously monitoring software, detecting vulnerabilities and reducing dependence on manual security operations.

For Microsoft, Project Perception represents more than another AI product. It reflects a strategic attempt to combine multiple leading AI models within a single platform while addressing one of enterprise customers' biggest concerns: balancing cutting-edge capabilities with manageable costs.

The company has not officially announced Project Perception, and details of the product remain based on reporting by The Information. However, if the launch proceeds as expected, Microsoft's latest offering could further intensify competition in the emerging market for AI-powered cybersecurity tools, where technology companies are racing to build systems that can outpace both human hackers and AI-assisted cyberattacks.