All it has to do is just recognize Anti-AI behavior as a health or security risk, that's it. There was an anti-crime AI that flagged Congress members too, so its not far fetched.