New AI Model Automates 90% of Vulnerability Detection Tasks

MAI-Cyber-1-Flash is an artificial intelligence model engineered to automate vulnerability detection and patching operations. Designed to integrate into the MDASH system, the model independently handles up to 90% of vulnerability detection tasks in the CyberGym environment. The remaining 10% of complex cases are delegated to more advanced models such as GPT-5.4. This tiered design optimizes computational resource usage by reserving more powerful models only for vulnerabilities requiring exceptional analysis. The architecture improves cybersecurity system efficiency by automating routine tasks while concentrating computational power where it is truly needed.