Microsoft unveils AI model for cybersecurity
Microsoft cybersecurity AI model spots risky code and finds vulnerabilities faster Boost threat detection, cut costs, and see how Project Perception helps fix flaws
Microsoft has introduced a new artificial intelligence model designed to identify risky areas in source code and help detect cybersecurity vulnerabilities. The company said the model, MAICyber1Flash, is intended to support its security tools and improve threat detection.
Microsoft said the model, when combined with OpenAI’s GPT5.4, performs better than competing security models on the CyberGym benchmark while operating at roughly half the cost. The system will be used in Project Perception, a set of AI agents for finding and fixing vulnerabilities, which is scheduled to begin public preview testing in early August.
The launch is part of Microsoft’s broader effort to strengthen its cybersecurity business after a leadership change earlier this year. The company is also emphasizing inhouse AI development as it looks for more efficient ways to deliver security products.