AVERI's Miles Brundage also shares these preemption concerns
Miles Brundage#20
Still deciding what I think about the overall merits of the new Obernolte-Trahan discussion draft, but I do have some concerns about the preemption provisions.
@ShakeelHashim the nature of the perfect compromise: everyone hates it equally.
Too narrow for Scalise, too broad for everyone else... odds of this going anywhere don't look good.
AVERI's Miles Brundage also shares these preemption concerns
Miles Brundage#20Still deciding what I think about the overall merits of the new Obernolte-Trahan discussion draft, but I do have some concerns about the preemption provisions.
Sunsetting preemption after 3 years (at the same time the affirmative provisions lapse) is better than permanent preemption for temporary affirmative provisions, but it's not like the state regulatory infrastructure would only snap back.
I also think the phrasing in terms "existing" common law remedies may preempt future evolution of common law tort doctrine as applied to AI, such as applying the abnormally dangerous activities doctrine or a vicarious liability theory.
It also cuts off state legislative efforts to update tort liability and impose liability insurance requirements, at least to the extent that they would apply to AI development as opposed to deployment. How does this apply to internal deployment?
As I said, I'm still evaluating the merits of the affirmative provisions to determine whether this is a trade worth making, but a few tweaks to the preemption provisions would go a long way to improving this bill.