Re: Coordinated Disclosure in the LLM Age

[Peter Gutmann <pgut001 () cs auckland ac nz>]

Jacob Bachmeyer <jcb62281 () gmail com> writes:

> You are correct here:  you should assume that any LLM will give a similar
> result to another person who asks a similar question.  In other words, LLM-
> discovered vulnerabilities should be considered already publicly known.

Not sure if this makes it better or worse, but I've found that Claude/Opus at
least gives different answers for the same query run over time.  If the thing
hasn't fallen out of context yet it'll tell you that we already did this one,
but otherwise it comes up with at least some different stuff each time.

Peter.