The GitHub repository "anthropics/defending-code-reference-harness" provides a reference implementation for autonomous vulnerability discovery and remediation using Claude. It includes tools for threat modeling, scanning, triage, and patching, along with a customizable scanning harness based on collaboration with security teams.

AI has transformed vulnerability discovery, but closing source code does not eliminate potential attack surfaces. Implementing continuous AI defense strategies is a more effective response to these threats.

AI agents are increasingly effective at identifying vulnerabilities in large software systems. Anthropic chose not to release the Mythos model due to concerns over its potential to discover dangerous security flaws.

N-Day-Bench evaluates the ability of frontier language models to identify real-world vulnerabilities disclosed after their knowledge cut-off dates. The benchmark features a standardized testing environment and monthly updates to test cases, focusing on the vulnerability discovery capabilities of large language models.