Microsoft has officially introduced the first specialized artificial intelligence model for cybersecurity tasks, named MAI-Cyber-1-Flash, as well as the new Perception agent platform. The presentation of innovative solutions took place at the company's profile event in San Francisco. The MAI-Cyber-1-Flash model is designed specifically to identify complex vulnerabilities in large-scale code bases and functions as part of a framework designed to detect and eliminate software defects. The Perception platform, in turn, uses AI agent teams to fully automate protection processes, including bug detection, prioritization, subsequent elimination, and deep integration with existing development systems.

The Perception platform architecture includes three specialized types of agent-based teams that provide an integrated approach to information security. The Red team simulates potential attack vectors by analyzing the behavior of intruders and identifying vulnerabilities that can be exploited by them. The Blue team is responsible for the prompt detection and ranking of existing problems according to the degree of criticality, while the Green team is directly involved in the development and implementation of methods to eliminate them. Microsoft's Vice President of Security, Hayet Gallo, emphasized that attackers are increasingly using artificial intelligence algorithms, so the new tools are designed to help corporate professionals fend off such threats with comparable speed and scale.

The new Microsoft software products will be available to corporate customers in a preview mode starting on November 3 of this year. This decision marks the entry of the technology giant into the rapidly growing market of specialized AI tools for cybersecurity. Earlier, similar initiatives were announced by key competitors: Anthropic introduced the Mythos platform to a limited number of partners as part of the Glasswing program, and OpenAI Corporation launched a similar solution through its own Daybreak program. This consolidation of efforts by leading developers confirms the formation of a steady trend towards automating information security processes using advanced language models.