Make rogue AIs more threatening

[Paineframe]

Because, as I just mentioned, not many traitors try to actively cooperate with a subverted AI, and trying to help out a traitor who you don't even know the plans of can be dodgy at best. Not only that, but some folks are fond of laws that specifically turn the AI into a murderbot and order them to purge all non-humans or whatever.

So cooperate with the traitor. Bolt doors and toggle APCs to support his activities. Or just make do with what little you can do. Nobody expects a lone AI to single-handedly murder the entire station, just like it's not able to do much else single-handedly (AI is very much a "helper", not a "doer") and you haven't really given a reason for why the AI should be good at killing people, other than "it isn't good at killing people".

The way I see it, it's a similar problem that the blob had (and arguably still has) as far as usefulness is concerned. If a rogue AI can avoid suspicion for long enough, build itself up in just the right manner, and get a network of informants going, it can be an absolute terror to the station. The problem is that the number of times this DOESN'T happen vastly outnumbers the number of times it does. If an AI slips up even once or gives off the slightest hint of being rogue, folks are going to grab the hacking tools and then it's game over unless a cyborg or antag comes to the rescue, which in my experience, rarely happens. Similarly to the earlier blob, the AI has very few methods of defending itself in the event of an attack unless it has backup.

Getting your laws reset isn't an attack, nor is it game over. It just means you're no longer allowed to kill people.