No, AI also needs to fail in similar ways as humans. A system that makes 0.001% errors, all totally random and uncorrelated, will be very different in production than a system that makes 0.001% errors systematically and consistently (random errors are generally preferable).