AI-found bugs aren't proving any easier to exploit despite the hype

I was under the impression the 'dynamic automation' of chaining these sorts of exploits was the real concern?