Microsoft announced new AI tools aimed at helping customers automate and streamline the process of identifying and mitigating security risks. The tools are part of a broader push to address the growing complexity of cyber threats. The company said the new models are built on its MAI-Thinking-1 platform and are designed to process over 1 trillion security signals daily. Microsoft described the tools as essential for managing the increasing scale and speed of cyberattacks. The company said the tools are currently in preview mode and should be closely scrutinized before use in production. Source: arstechnica
The first tool, AI-Cyber-1-Flash, is a security model specifically trained to identify and fix software vulnerabilities. It is integrated into MDASH, a multi-model agentic scanning harness introduced in May. MDASH combines 100 security-trained AI agents to discover exploitable bugs in applications. Microsoft claimed that MDASH with AI-Cyber-1-Flash received a 96 percent score on CyberGYM, a standard benchmark test, outperforming competitors like Anthropic’s Mythos, Google Gemini, and OpenAI GPT. The new MDASH offering is said to cost half as much as its predecessor. Source: arstechnica
Microsoft also announced Project Perception, a platform of specialized AI agents that perform red-, blue-, and green-team functions for vulnerability detection, investigation, and remediation. The platform selects models based on task effectiveness and customer cost. Microsoft said Project Perception is designed to handle 90 percent of tasks at lower costs than similar platforms from competitors. The company emphasized the need for security teams to adapt to the rapid evolution of cyber threats, as AI accelerates both attack speeds and the complexity of digital environments. Source: arstechnica
