I assume avoiding this is the entire reason OpenAI is testing exactly that.

The hacks can be stopped, the virus can be a whole another level