System and method for triaging vulnerabilities by applying bug reports to a large language model (llm)

This invention describes a way to automatically prioritize software vulnerabilities. It works by feeding raw, unstructured text from bug reports into a smart computer program, like a large language model. This program then generates several scores for each vulnerability, indicating how likely it is to be exploited, attacked, or become a publicly known threat. These scores help security teams quickly decide which vulnerabilities need urgent attention. The claims specifically include generating a third score for public exposure and using user feedback to improve the system.