Microsoft touts cost-saving AI model for cybersecurity

Key Points:

• Microsoft introduced a new AI model, MAI-Cyber-1-Flash, designed to identify cybersecurity vulnerabilities, marking its first major effort to revitalize its cybersecurity business under executive Hayete Gallot.
• When combined with OpenAI's GPT-5.4, MAI-Cyber-1-Flash outperforms competitors like Anthropic's Mythos 5 and Google's 3.5 Flash Cyber on the CyberGym benchmark, offering leading performance at half the cost.
• The model will be integrated into Project Perception, a suite of AI agents for detecting and fixing security weaknesses, which will enter public preview on August 3.
• Microsoft aims to leverage specialized AI models and in-house training to improve cost efficiency and outcomes in cybersecurity, while addressing the growing threat posed by attackers using generative AI.
• Despite a 19% drop in Microsoft shares this year amid concerns over AI competition, analysts remain optimistic, and the company continues to expand AI-driven security tools, including the Security Copilot and integrations with non-Microsoft products.