How Visa Secures Trillions Using AI Agents & Open-Source Harnesses

AI security requires a shift toward autonomous, machine-speed defense mechanisms to counter the rapidly shrinking window between vulnerability discovery and exploitation. Subra Kumaraswamy, CSO of Visa, details the development of the open-source Visa Vulnerability Assessment using Agentic Harness (VVAH), a tool designed to guide models in identifying and remediating vulnerabilities across complex enterprise stacks. By moving beyond siloed, deterministic security tools, organizations can leverage reasoning engines to synthesize cross-domain signals, enabling faster threat detection and response. Visa’s strategy prioritizes "Mean Time To Adapt" (MTTA) as a vital metric, ensuring that patches are not only identified but validated and deployed at scale. This approach emphasizes that security teams must adopt a developer mindset, utilizing automation and agentic frameworks to maintain trust and resilience in an increasingly AI-driven threat landscape.