Microsoft AI Security: Project Perception Fights Cyber Threats

Microsoft has unveiled its first specialized AI security model for cyber defense, called Project Perception. This new system aims to counter the rise of AI-driven cyber threats. The company states that traditional security tools can no longer keep pace with autonomous systems that generate exploits faster and scale campaigns further. Project Perception continuously analyzes risk across an organization's identities, endpoints, applications, data, cloud services, and AI systems. It uses three groups of specialized AI agents: red team agents probe for vulnerabilities, blue team agents investigate risks, and green team agents handle remediation. Human defenders still set objectives and must approve high-impact actions. The system employs a multi-model architecture, using models like MAI-Cyber-1-Flash, which is now integrated into MDASH. This updated configuration scored 96% on CyberGym, an industry benchmark, outperforming Mythos by 12 percentage points and nearly halving costs. Project Perception will integrate directly into Microsoft's security products, starting with Microsoft Defender, to turn findings into corrective actions. This system enters public preview on August 3 with consumption-based pricing. This development could help organizations keep pace with increasingly sophisticated and automated cyber attacks.