Visa used Mythos to hunt for bugs in its own payment network, then open-sourced the harness that made it possible

Visa used Anthropic’s Claude Mythos to probe its global payment network and found chained vulnerabilities that could be exploited in complex ways.

The company then open-sourced the Visa Vulnerability Agentic Harness, the system it built to manage AI-driven security testing.

Visa also published a white paper detailing the harness architecture and the security controls used during testing.

The case highlights how frontier AI can help uncover risks in critical payment infrastructure and how defenders are adopting agentic tools, not just scanners, to harden large networks.