I suspect we are near the point where the only way to keep up with LLM discovery and exploitation (and creation?) of software and hardware vulnerabilities is with other LLMs, unfortunately.

I don't know what could be done about the parallel LLM discovery and exploitation of human vulnerabilities.