Key facts
- Microsoft launched MAI-Cyber-1-Flash, its first AI model for cybersecurity.
- The company also released Perception, an AI platform for automating security workflows.
- MAI-Cyber-1-Flash achieved a 96% score on the CyberGYM benchmark, surpassing competitors.
- Perception utilizes AI agents for vulnerability detection, risk assessment, and remediation.
- Microsoft claims the new tools offer significant cost reductions compared to existing solutions.
- The tools will be available in preview starting November 3.
Microsoft has launched its first artificial intelligence model specifically designed for cybersecurity, named MAI-Cyber-1-Flash, alongside a new AI-powered cybersecurity platform called Perception. The company announced the developments, positioning its new tools as significant advancements and direct competitors to offerings from Anthropic, Google, and OpenAI.
MAI-Cyber-1-Flash is built to identify challenging vulnerabilities within complex codebases and is integrated into Microsoft's MDASH harness for software vulnerability identification and remediation. Microsoft claims this model is more powerful and cost-effective than existing competitor models, citing a 96 percent score on the CyberGYM benchmark, which is 12 points higher than Anthropic's Mythos and also surpasses Google Gemini and OpenAI GPT. The new MDASH offering costs half as much to use as previous versions.
The Perception platform is designed to deploy teams of AI agents to automate and assist with various security workflows, including bug identification and remediation. It can also integrate with MDASH. Microsoft stated that Project Perception is designed to perform 90 percent of tasks for lower costs than similar platforms from competitors, allowing customers to reserve more expensive alternatives for the remaining 10 percent of tasks. The platform selects models based on task effectiveness and customer cost, informed by ongoing research and benchmarking.
These new tools respond to a shift in cybersecurity where AI accelerates attacks, necessitating advanced defensive capabilities. Microsoft noted that security teams struggle to keep pace with emerging threats due to the complexity of digital environments and the need to piece together vast amounts of data. The tools, currently in preview, are being introduced with a call for caution and thorough evaluation before production use, balancing the risks of adopting AI agents against the threat of not using them.