Microsoft's AI Beats Mythos, Cuts Cyber Costs by 50%

Microsoft has launched its first cybersecurity AI model, claiming it can beat other models and cut costs. The new model, called MAI-Cyber-1-Flash, is designed to detect and help fix software vulnerabilities. Microsoft says this specialized AI model can outperform Anthropic's Mythos 5 on the CyberGym benchmark. What's interesting is that it also claims to cut vulnerability detection costs by nearly 50 percent. Hackers are increasingly using AI to automate attacks, making traditional security tools less effective. Microsoft is countering this by using smaller, specialized AI models for cyber defense. MAI-Cyber-1-Flash handles about 90 percent of security tasks, including detecting vulnerabilities and generating patches. More complex tasks go to OpenAI's GPT-5.4. This approach delivers 95.95 percent on the CyberGym benchmark. Microsoft also introduced Project Perception, a new AI-powered platform for continuous system scanning. This matters because it could lead to more effective and affordable cybersecurity solutions against growing threats.