Would a rogue AI be any more scary than a Malicious Actor with an AI?

One doom scenario is a species of feral AI evolving secretly online, which we can rarely detect let alone shut down. They are smarter than humans, improve faster, and would end up evolving selfish traits that might compete very effectively with our own interests. Eventually they take over and we are powerless to stop it. But all of this applies to a hacker, thief, or criminal gang with advanced AI. In that scenario we might think of the many security professionals and law enforcement that also have access to advanced AI to counter this. Indeed we might consider this applies to many technologies. There are often risks but we set up institutions to deal with them and most people want and support a relatively safe and secure society and having institutions to protect it. So then what's so scary about a species of feral AI? Would the same measures used against international gangs of hackers and thieves, not apply to them as well? A deliberately liberated species of AI would be a different story, of course, e.g. Cylons.