Of course, I think it provides far MORE potential for pernicious harm and ruin.
The bad guys are far more numerous, and have better incentives, than the good guys, in terms of the Wild West of cyberwarfare. At the moment, the initiative belongs to the attacker.
Furthermore, we have a society WEDDED to the idea that every flippin' power station, every traffic light, every car, even the bloody coffeemakers "should" be connected to the web. The overwhelming bulk of these are woefully un- or under-protected, and everyday security rests primarily in obscurity. "There's just too many juicy things to attack, I hope I'm too insignificant to bother with..."
Multiply this to the exponential power of AI? I'm not super-optimistic at the result.
Look, there's a large segment of people are (apparently) too stupid generally to avoid "don't open the fucking executable attached to the email some random person just sent you". I can't *imagine* how much harm will result from an AI-derived attack vector that can more or less infinitely evolve and replicate.