"making AI algorithms subject to verification and transparency;"
That will ban very large numbers of algorithms which simply cannot be verified - they've been generated by evolving, on an enormous data set.
You might be able to test them for built in bias (though you have to suspect it before you can test for it) but you can't verify them or explain how they work.
Whether banning such algorithms is a good thing is a different question.
Biting the hand that feeds IT
