Microsoft launches MAI-Cyber-1-Flash cybersecurity AI model; platform enters public preview on August 3

Microsoft's MAI-Cyber-1-Flash cybersecurity AI model, launched on August 3, aims to revolutionize security operations by automating approximately 90% of routine tasks. This model, integrated with OpenAI's GPT-5.4, achieves a remarkable 96% on the CyberGym benchmark, outperforming competitors like Anthropic's Mythos by 12 percentage points at half the cost.¹²

"We have world-leading performance at 50% of the cost," said Mustafa Suleyman, CEO of Microsoft AI. The model is designed to handle the bulk of security work, allowing GPT-5.4 to focus on more complex issues. This shift is expected to significantly reduce the time spent on security tasks, with Weston noting, "We've gone from this taking hours and hours of manual work... and in minutes, we have a fix for all of this."

The platform's launch comes as Microsoft continues to bolster its cybersecurity offerings, which generated over $20 billion in revenue in 2023. Gallot, overseeing the security division, emphasized the importance of using AI to defend against AI threats, stating it provides enterprises the capability to "defend against AI with AI at the scale and speed that the attackers have."