A dangerous tipping point? AI hacking claims divide cybersecurity experts