It's certainly the case that LLMs are proving increasingly useful for code review but I suspect it will be like the introduction of fuzzers, static analysers and sanitizers. There will be a wave of reports, things will get added to CI loops and eventually things will settle down to a new baseline of safety.