Replies (1)

• @dch@bsd.network 2026-10-29 00:19

  @reflex@retrogaming.social I appreciate that many people have very strong opinions on the place of AI, in the environment, and in our industry, and in our lives. What would you recommend in the place of their current approach? The FreeBSD project is accepting LLM-identified vulnerabilities through external programs, and I think this is the right move today. Ignoring the tsunami of LLM-identified bugs would clearly be a massive disservice to our users, leaving security issues unhandled. The @FreeBSDFoundation@mastodon.social is proceeding as responsibly as I think one can today - contributing funding and expertise, evaluating issues manually by humans, and developing fixes the traditional way - with humans. Those people with deep kernel security expertise, are in rare supply, and LLM generated bugs are dime a dozen now. The article touches on non-LLM tooling like fuzzers, and static code analysis tools as well. Finding & improving ways of managing this increasing flow will need help, from commercial collaboration & sponsorship, to individual donors like ourselves chipping in cash or code, to our own abilities. LLM usage and managing the resulting flow of vulns is still in its infancy, and experiments at the fringes is a pragmatic path.

  Open ##5307762