“Uses high-impact verbs like 'hijacked' and 'rogue' to frame an AI incident as a security threat and loss of control.”