Microsoft launches AI model for cheaper code security scanning

Microsoft announced MAI-Cyber-1-Flash, an artificial intelligence model built specifically to find security flaws in source code. The model will operate inside Project Perception, a group of AI agents designed to detect and repair vulnerabilities, working alongside OpenAI's GPT-5.4.

According to Microsoft's testing, the combined system outperformed competing security models on the CyberGym benchmark while operating at about half the cost. Project Perception will enter public preview on Aug. 3. The tools can recommend code changes and apply fixes with customer approval. They can also connect with products outside Microsoft's ecosystem, giving enterprises flexibility in integrating the system with existing security infrastructure.

This launch represents Microsoft's first major cybersecurity initiative since Hayete Gallot returned in February to lead the business. The company cited the accelerating pace and sophistication of cyberattacks as a driver for the investment, noting that generative AI has lowered the cost of both attacking and defending software systems. Microsoft's cybersecurity business generated annual revenue exceeding $20 billion（約3.2兆円） in 2023, though the company has not updated that figure since. Investors will be watching adoption of Project Perception, the cost savings it delivers, and future cybersecurity revenue growth—while also observing Microsoft's broader strategy of investing more heavily in its own AI models as questions persist about reliance on OpenAI and rising infrastructure spending.

Microsoft's move into specialized AI-driven cybersecurity marks a strategic pivot as the company grapples with two overlapping pressures: the rising cost and sophistication of cyberattacks, and investor concerns about its reliance on OpenAI partnership and infrastructure spending. By building MAI-Cyber-1-Flash as a lower-cost alternative to existing security models—one that achieves better performance at roughly half the cost—Microsoft is positioning itself to expand its cybersecurity business, which generated over $20 billion（約3.2兆円） in annual revenue as of 2023 but has not been updated since that time. The timing aligns with leadership continuity: Hayete Gallot's return in February to lead the cybersecurity business suggests organizational commitment to this sector.

Project Perception's architecture—combining Microsoft's new model with OpenAI's GPT-5.4 and positioning it as an agent-based system that can integrate with non-Microsoft products—suggests Microsoft is hedging its bet on exclusive OpenAI dependence while building compatibility and stickiness through interoperability. This approach may appeal to enterprises with heterogeneous security tooling environments, lowering barriers to adoption.

No comments yet. Be the first to share your thoughts!