It's pretty simple, really. This would be a non-issue if you programmed your cyborg to go pick up milk from the store and it started handing out $20s to strangers in the dairy aisle. Obviously that's no fault but your own.
Why is it different for a webserver?
It is massively unfair to expect someone to make assumptions about the intent of a remote system, programmed, configured, and deployed by people they have never met or communicated with, in order to avoid criminal liability.
I'm not sure what the problem is here. It seems like simple common sense to me.