In the face of Federal US failure to actually regulate AI, the states are beginning to act on their own.

We have previously posted about the cautious pace of US Federal regulation of AI to limit risk and fraud, and speculated that such a hands-off approach to new technology is an American characteristic: let new tech flourish and do not interfere, as in the long run it will drive US primacy in the business and scientific worlds.  We have also noted that the EU has been far more aggressive in putting current regulation in place.

Noting the failure of the Feds to take action, the National Conference of State Legislatures recently published a report discussing what AI regulation should look like in financial services, healthcare and education. In fact, fourteen states have already enacted some legislation relating to control of AI, and another 15 have introduced legislation, much focused on setting up research committees to work on legislative initiatives.

Among areas of focus in the report: preventing use of AI in hiring due to fear of bias in the data relied upon; monitoring algorithms utilized in making decisions in regulated areas including education, housing and banking; protecting intellectual property rights, which can be trampled when AI sucks up information that is proprietary and incorporates it into machine analyses; establishing a common vocabulary for states to use in enacting legislation so there can be understanding and coordination.

This last suggestion of a common vocabulary, which is facially logical, discloses the fundamental weakness of hiving separate laws in a multiplicity of jurisdictions: AI crosses borders continually, companies developing and using AI do not operate in or impact just one state, and for lawyers and businesspersons trying to comply with multiple laws is expensive and also not wise.  What is needed is National legislation that preempts the field, replaces state laws and creates a uniform national standard.

Now the Federal Congress can let a hundred flowers bloom at the state level, evaluate impact and wisdom of these various schemes, and then adopt uniform laws that supersede inconsistent state regulation, having used state law temporarily as a laboratory for evaluating what is wise and what is not.  That leave chaos at the beginning and resistance by states to abandon what their own legislatures have enacted and thus feel is simply too good to lose.

But between a propensity to move slowly in Federal tech regulation, and the complex problems presented by regulating AI, we may have no choice but to end up with a Balkanized AI regulatory landscape.