It certainly is unsafe for their AI training corpus. Win / win if they can also lie about protecting children as a motivation.