Microsoft has introduced two new AI-driven security platforms, MDASH and Project Perception, designed to automate vulnerability detection and network defense.
Key facts
- •MDASH scored 96 percent on the CyberGYM benchmark, outperforming Anthropic, Google, and OpenAI models.
- •Project Perception automates red-, blue-, and green-team security functions.
- •Microsoft claims Project Perception can perform 90 percent of security tasks more cheaply than competitors.
- •The new tools are currently available in preview mode.
- •Microsoft stated the tools address the increasing speed and scale of AI-driven cyberattacks.
Microsoft announced two new AI security tools on Monday intended to help organizations defend against cyberattacks. The company introduced MDASH, an updated offering, and Project Perception, a collection of specialized AI agents designed to perform red-, blue-, and green-team functions.
By the numbers
MDASH Performance and Cost
Microsoft reported that its MDASH tool, utilizing MAI-Cyber-1-Flash, achieved a 96 percent score on the CyberGYM benchmark test. This rating is 12 points higher than Anthropic’s Mythos and also surpasses performance scores from Google Gemini and OpenAI GPT. Additionally, the new MDASH offering is priced at half the cost of its predecessor.
Project Perception Capabilities
Project Perception uses specialized AI agents to identify vulnerabilities, assess risk, and implement corrective actions. The platform automatically selects models based on task requirements, balancing effectiveness and customer cost. Microsoft claims the tool can handle 90 percent of security tasks at a lower cost than competing platforms, allowing users to reserve more expensive alternatives for the remaining 10 percent of work.
Advertisement
This article was independently rewritten by ManyPress editorial AI from reporting originally published by Ars Technica.
