He's not saying don't build anything. He's saying think about the potential for misuse (and implying, take reasonable steps to prevent it). This seems completely sensible to me.
I am aware I cannot think of every bad way a service could be misused . I also am aware that there is always a way to misuse a service. Therefore, in order to prevent a service I use from being misused, I must not build anything.
Obviously that conclusion is wrong, so one of the premises must be wrong. Specifically, liability should be on the abuser not on the service.