Anthropic releases open-source framework for AI-driven code vulnerability detection

Anthropic published a reference harness for using Claude to identify security vulnerabilities in code. The framework is designed to help developers integrate AI-assisted code scanning into their workflows without building vulnerability detection from scratch—useful for small teams that can't afford traditional security audits.