New National AI Plan will Defer to the Market, Pare Back Safety Regulations

Under an AI action plan released last month, the US will alter course on AI regulation.  The outgoing Biden administration limited export of AI chips outside the US, and instituted rules designed to increase safe development and deployment of AI.  The Trump administration has repealed much of the Biden AI agenda and will allow a much more laissez-faire approach both to AI exports and to governance matters regarding its use.  The US will also, under the new plan, use federal land and relax environmental laws to allow rapid development of data centers to house AI operations.  

WHY IT MATTERS

The US has no national regulation regarding AI, although some states have made small steps toward regulating it.  Those state laws (and other proposed state laws that have yet to pass) generally have focused on protecting humans from potentially harmful use cases involving AI (such as implementing disclosure requirements, internal risk assessments and reporting, or prohibiting or curbing the use of AI under certain circumstances).  In their form and substance, these laws tend to take a risk-based approach to AI regulation: the greater the risk of harm to humans, and the worse the potential harm, the more regulation.  This mirrors the approach taken by the EU, which is the largest trade partner of the US, in its bloc-wide AI Act.  Although the Trump administration takes a very different approach from the Biden administration, this is not purely a political (red v. blue) issue: California has considered and rejected several AI proposals, whereas Texas has passed the most comprehensive AI law in the US to date.  Other states are likely to consider AI regulation in the near future.  

As the nation's new AI plan comes to fruition, it will have wide-ranging impact.  AI chip makers are likely in for increased sales if they are allowed to sell relatively freely overseas.  Consumer watchdogs are likely to push back on looser governance standards that could leave AI decision-making in the hands of companies that don't have to disclose how they use AI and what its impact is.  National security concerns are likely to become part of the conversation as well: it is not clear how allowing the sale of AI chips in China differs from the sale of personal data to China or the use of Chinese-owned social media in the US (both of which have been banned in the last two years).  

Because the AI industry is so nascent, it is hard to project how a relatively hands-off scheme will work, both inside the US and as a matter of trade policy.  Companies using AI technology now would do well to document their decision-making process and the legal basis for their decisions, in case standards change/tighten over time.