I don't think the copy.fail people understood the issue they found, as is evident by the heavy focus on AF_ALG/aead_algif, which is essentially "innocent" as we're seeing here.
I think LLMs are great for vulnerability discovery, but you need to not skimp on the legwork and understanding what even you just found there.
