Bitcoin Core has created a new category of vulnerabilities described as “easily discoverable,” which includes flaws that can be identified with simple prompts sent to large language models. Fixes for these flaws will be backported immediately to all actively supported major versions and accompanied by a security advisory published without an embargo. Low-, medium- and high-severity vulnerabilities will nevertheless retain their existing disclosure schedules, while critical flaws will continue to be handled on an ad hoc basis. The decision responds to the growing number of reports facilitated by AI tools, but no vulnerability has yet been assigned to this category, and the definition applies regardless of the tool used to discover the flaw.
Source: Read the original article
