Microsoft Says Its New Cybersecurity AI Beats Industry Leaders at a Fraction of the Cost
Microsoft has a new AI cybersecurity model called MAI-Cyber-1-Flash, and whzu combined with the MDASH agent protection system and the OpenAI GPT-5.4 model, it outperforms Anthropic Claude Mythos 5 by 12 points in the main benchmark. The security product is designed to “use AI to protect against AI,” Microsoft said.
The combination of MAI-Cyber-1-Flash and MDASH – launched in May – is called Project Perception, and goes into public preview in Aug. 3, built directly into Microsoft Defender. It will be gradually rolled out across all Microsoft Security products.
According to benchmarks posted by Microsoft on Monday, the combination scored 96% on CyberGym, compared to Mythos 5’s 84%.
Pricing is based on usage, measured by the number of security computing units you use. As the AI agents do scenarios, they use SCUs, so the more work is done, the more you pay – but Microsoft says the cost savings is about 50% of the current MDASH configuration on the market now.
Speaking at a Microsoft forum on Monday morning, Mustafa Suleyman, CEO of Microsoft AI, explained the handover process between MAI-Cyber-1-Flash and GPT-5.4.
“MAI-Cyber-1-Flash handles about 90% of the queries. It detects the vulnerability, catches it, sends it and proves that it was a valid and correct solution. Then it defers about 10% of the queries to GPT-5.4, which is obviously a large model, “and 10 “Together, since the models complement each other, in fact they can not only deliver better performance than all other combined models – they do so at a cost 50%.
Suleyman called the CyberGym’s standing effect “a very impressive result.”
It follows the launch of Anthropic’s Claude Fable 5 last month, the first publicly available model from the Mythos family. At the time, Anthropic said that the Mythos was so good at finding cybersecurity flaws that it could break the Internet if used unchecked – and Anthropic was forced to back off the launch of Fable 5 and Mythos 5 within days because the US government said it knew how to “jailbreak” the model and bypass restrictions. When Mythos was first announced, it was only released to select government agencies and technology experts.
“Microsoft has long been a trusted custodian of the world’s most valuable, most valuable government and business data for decades. We’ve accumulated an incredible amount of data over that time,” Suleyman said Monday. “It’s that data combined with the expertise we have of world-class cyber security experts at the company that we’ve been able to really drive this integrated model.”
See the full history
