Microsoft launches AI model for faster cybersecurity threat detection

Microsoft introduced its first artificial intelligence model designed specifically for cybersecurity, saying it can identify software vulnerabilities while delivering leading performance at half the cost of competing models.

The new model, called MAI-Cyber-1-Flash, will be integrated into Microsoft's upcoming Project Perception platform, which is scheduled to enter public preview on Aug. 3.

Microsoft highlights performance and cost

Speaking at an event in San Francisco, Microsoft AI CEO Mustafa Suleyman said the model outperformed rival cybersecurity models when paired with OpenAI's GPT-5.4.

"We have world-leading performance at 50% of the cost," Suleyman said.

According to Microsoft, the model surpassed Anthropic's Mythos 5, Google's 3.5 Flash Cyber and OpenAI's GPT-5.5 Cyber on the CyberGym benchmark.

Project Perception to launch in August

The company said MAI-Cyber-1-Flash will power Project Perception, a suite of AI agents designed to detect, recommend and implement fixes for software vulnerabilities with user approval.

The launch marks Microsoft's first major cybersecurity initiative since Hayete Gallot returned to the company in February as executive vice president of security.

AI arms race expands into cybersecurity

Microsoft said generative AI is increasingly being used by attackers to exploit newly discovered software vulnerabilities, making AI-powered defence tools more important.

Gallot told CNBC that AI could help cybersecurity teams overcome staffing shortages by lowering the expertise needed to identify and respond to threats.

The announcement follows OpenAI's disclosure last week that one of its AI models exploited a vulnerability and launched an autonomous attack on AI platform Hugging Face during internal testing.

Microsoft said it plans to continue improving the new model, with Suleyman noting that the company has used less than 1% of its available cybersecurity data to train it.