Microsoft has significantly expanded its artificial intelligence (AI) cybersecurity portfolio with the introduction of its first dedicated AI security model, MAI-Cyber-1-Flash, and a comprehensive new AI cybersecurity platform named Perception. Unveiled at an event in San Francisco, these new tools are designed to provide businesses with advanced capabilities to detect and neutralise increasingly complex cyber threats, particularly those leveraging AI.
MAI-Cyber-1-Flash is described by Microsoft as a model specifically engineered to pinpoint challenging vulnerabilities within intricate codebases. It is built to integrate with MDASH, Microsoft's existing harness for identifying and remediating software vulnerabilities. Alongside this, the Perception platform is designed to deploy teams of AI agents – categorised as red, blue, and green teams – to automate various security tasks. Red teams simulate potential attacks, providing crucial context on threat actors and likely exploits, while blue teams focus on detecting and triaging existing bugs. Green teams are then responsible for taking corrective actions to fix these vulnerabilities.
Microsoft asserts that MAI-Cyber-1-Flash is both more powerful and more cost-effective than rival models, citing its strong performance on established AI cybersecurity benchmarks. Mustafa Suleyman, CEO of Microsoft AI, highlighted the model's superior results, stating, "We have MAI-1 Cyber Flash binded with GPT 5.4 inside of the MDASH harness — which beats out Gemini, GPT 5.5 Cyber, GPT 5.6 Sol, and Mythos 5 on Cyber Gym, which is the primary benchmark that we all use. The golden benchmark." He added that the company is shipping these innovations into production immediately.
The launch comes as cybercriminals increasingly harness AI to develop more sophisticated and rapid attacks. Hayete Gallot, Microsoft’s Vice President for Security, emphasised that Perception offers enterprise defenders a way to "defend against AI with AI at the scale and speed that the attackers have." Dave Weston, lead engineer for Perception, underscored the platform's efficiency gains, noting that tasks that once required hours of manual work from multiple specialists can now be accomplished in minutes, from discovery and prioritisation to detection, posture fixing, and even code remediation.
For UK businesses, these advancements could represent a significant upgrade in their defensive capabilities. As the digital landscape evolves, the UK's Information Commissioner's Office (ICO) and the broader regulatory environment, including the EU AI Act's implications for businesses operating within the bloc, will continue to shape how AI-driven cybersecurity tools are deployed. Expert commentary suggests that while AI offers immense opportunities for strengthening cybersecurity, it also introduces new ethical and operational considerations. The new tools are expected to be available in preview from November 3, 2026, entering an already competitive market that includes solutions from companies like Anthropic and OpenAI.