One day he says AI is frightening and must be regulated, one other day he says I don't want AI regulations...
He didn't say he doesn't want them. He said AIs, for a lot of reasons, may not be able to operate within the confines of potential regulations. For a simple example, they may be trained on work that was not properly licensed in the EU. That's only one compliance challenge in one dimension, but there are a large number of dimensions in which AIs might fall outside the mandated parameters.
Obviously people will try, but there are reality based technical limitations.