The public behavior loophole is likely protected by first amendment. While it is true that common rule can be used as criteria for denying grant applications, it has potential of having chilling effects on number of experiments. Consider recent imagenet type challenges which rely on images collected from public datasets such as flickr. Going ahead each study will require an IRB approval with enormous breaucracy burden leaving only corporations without any restrictions to do this type of research. There is a strong precedent in IMS health vs Sorrel by Supreme Court against speaker based restrictions on speech.
Finally I am sick of hearing Anonymization != De'identification it's tired trope where there are barely any solutions put forward other citing a bunch of papers and calls for more breaucracy.
There is an interesting article by Dr. Atul Gawande on IRB issues when implementing a simple checklist that reminded doctors to wash their hands.