in testing, smaller ai models can now autonomously validate potential security vulnerabilities, though this may produce some false positives when used at scale.