Trump’s AI Executive Order Carries Benefits for Providers, Users

In many ways, a single national standard on artificial intelligence would benefit both providers and users of AI technologies. President Donald Trump’s Dec. 11 executive order seeks to promote AI development in the US by freeing AI companies from excessive and cumbersome state regulation.

The executive order, titled “Ensuring a National Policy Framework for Artificial Intelligence,” cited the following:

• Individual state laws create a patchwork of overlapping and possibly conflicting regulations.
• State laws may require or cause AI companies to embed ideological bias within their models (suggesting, for example, that Colorado’s ban on “algorithmic discrimination” resulting in “differential treatment or impact” may distort AI output).
• Individual state laws may affect AI-related transactions outside those states.

It’s costly for AI providers—especially startups—to remain current with 50 different regulatory schemes on the same topic. Although the process can be streamlined to an extent, it’s difficult to create standard templates and compliance programs that would account for the varieties of state laws. The lawyers and consultants who are needed to track and interpret those regulations add time and expense to any AI business.

Compliance, training, reporting, and technology requirements all create barriers to growth and innovation by early-stage companies that can’t readily bear the financial burden.

It’s also expensive and cumbersome for a provider (or its lawyers) to draft terms of use, consumer disclosures, and similar materials in a way that covers multiple states’ legal requirements while remaining understandable to a typical reader.

And different states enacting legislation to reflect their own public policy choices imposes a technological burden, where the AI provider must determine the state law governing each user and build models and algorithms to account for the unique policies of each state.

The European Union’s General Data Protection Regulation gives a useful analog to the proposed national regulation described in the executive order. The GDPR has become a useful reference standard since its enactment in 2016, once businesses adjusted to its requirements. It provides common definitions and obligations and a generally consistent set of privacy practices and documentation.

The GDPR benchmark is now an accepted body of expectations for businesses and consumers in the EU for transparency and accountability. Similarly, a national standard for AI tools may create a shared understanding and level of comfort among all AI providers and users that can be a competitive advantage.

A national AI standard likely would enhance consumer trust and confidence in AI tools, mitigate regulatory uncertainty, and assure continued innovation and investment from AI providers.

The 2007 amendments to the Federal Trade Commission’s Franchise Rule are an example of a national regulatory scheme designed to make franchise disclosure more relevant and uniform—and to streamline the disclosure and compliance burden on franchisors.

Because AI permeates virtually every aspect of interstate commerce, there may be ancillary state laws that affect AI, such as the use of AI tools in education, employment, health care, or credit decisions. This raises the possibility that a state may be accused of injecting ideological bias (to use the executive order’s words) into an AI tool.

That possibility should be preempted, according to the executive order—even though the same result without the use of AI technology generally wouldn’t necessarily be preempted by federal law.

Under the Constitution’s Supremacy Clause and Congress’ authority over interstate commerce, federal law will control if it specifically provides that it preempts state law, if compliance with both federal and state law isn’t feasible, or if state law unduly burdens interstate commerce. The edges of permissible state regulation and controlling federal law often blur, and the courts will have the final say on which authority prevails with AI.

As a policy statement, the executive order will have limited impact unless Congress enacts legislation on the issues it covers. It’s said that all policy is politics, and an act of Congress is likely given the strong lobbying presence of larger AI providers, the ideological goals of other interest groups, the amount of capital flowing to creating and deploying AI technology, and a need to maintain global competitiveness.

In anticipation of congressional action, users of AI should:

• Take advantage of the resources available from their trade associations or similar groups, including their lobbying power and collection of information.
• Monitor the National Institute of Standards and Technology’s website, which is a gold mine of information about AI standards and research, risk management, and other relevant topics.
• Seek protection in the terms of use or subscription agreements for AI tools by requiring the AI provider to warrant that its tool complies with applicable laws (defined broadly to include federal, state, and international laws); warrant that the AI tool is regularly vetted for bias; and indemnify the user for claims and damages caused by the design of the model and algorithms (but excluding any claims or damages caused by user input or the use of output).

This article does not necessarily reflect the opinion of Bloomberg Industry Group, Inc., the publisher of Bloomberg Law, Bloomberg Tax, and Bloomberg Government, or its owners.

Author Information

Matthew I. Hafter is partner at Thompson Coburn, advising founders, boards of directors, startup/venture-backed and mature companies, and investors in a broad set of transactions.

Write for Us: Author Guidelines