Spurred by recent executive orders concerning administrative agencies, how to conceive of “independent” regulatory agencies under the U.S. Constitution is front and center. This is about more than cost-benefit analysis, whether a federal regulatory agency acted outside its authority, or cost savings. This is about the separation of powers between Congress under Article I and the president under Article II within our constitutional structure. Do independent administrative agencies perform legislative functions, executive functions, or both?
The Supreme Court has spoken to this as far as back as the 1935 Humphrey’s Executor case. It’s expected that the court will address this question anew, determining the reach and interplay of the executive and legislative branches in light of the administrative state’s growth over decades.
Wherever the Supreme Court lands, one thing is certain: Whether a regulatory body falls under Article I, Article II, or some mixture, the practical challenge for regulators is to do their best to advance the public interest.
A role for AI in regulation
Getting regulation right is not easy. Today, federal regulatory agencies enjoy a singular opportunity to aid their performance using artificial intelligence, which is increasingly capable of reasoning and completing complex multistep tasks.
AI should be further integrated into the regulatory process. AI is well-suited to generate key insights for regulators and help them perform and streamline countless regulatory responsibilities.
What we have in mind begins with expressing regulation as computer code—that is, encoding regulatory requirements as computational representations in AI systems that leverage large language models to interpret and apply regulation, taking into account laws and regulations, guidance from regulators, industry practices, and court opinions. Many possibilities flow from there. Here are two promising options for regulators to explore.
Unlocking insights that improve regulation
First, regulation should be designed to meet the government's goals without imposing undue costs on businesses and individuals. That is easier said than done because it is difficult to identify the consequences of a regulation when it is being crafted. Government-hosted AI systems could be developed to evaluate which conduct likely would violate (or comply with) proposed regulation.
Regulators could use AI to run multiple scenarios against regulatory requirements to better predict how a proposed rule—before it is finalized—might affect people’s behavior and business activities. For example, if a regulator proposes to mandate certain disclosures, AI could help assess which disclosure content likely would (or would not) comply with the proposed mandates, along with how different audiences may react to different information provided to them.
With insights unlocked by AI, regulators could help ensure that a proposed regulation is likely to do more good than harm if adopted—or decide how the regulatory proposal should be revised to be more effective and less costly. Scenario analyses like this are not practical at scale with wholly manual processes, but they are with AI.
Promoting innovation, creativity, and entrepreneurship
Second, a lack of regulatory clarity, or a lengthy licensing or permitting process when a governmental approval is needed to do business, can hinder innovative goods and services from being brought to market—or new businesses from launching. As a result, society can lose out on things that could improve our lives.
To accelerate progress, regulators could deploy AI agents as a form of automated pre-clearance providing an initial check of whether a business’s proposed activities comply with regulatory requirements. Although a regulator’s human intervention and judgment still may be needed sometimes, businesses could interact with a regulator’s hosted AI agent by inquiring if certain conduct comports with a regulation encoded as a computational process into the AI system. In short order, the AI agent could respond indicating (with an explanation) whether the activities are approved, more information is needed, the activities raise concerns, or someone at the regulator needs to evaluate. As to a license or permit request, an AI agent could perform a first-level assessment of whether the application meets the requirements and, if not, explain.
By incorporating AI in these ways, regulators could provide more timely and efficient guidance and, in some cases, permissions to the marketplace—without frustrating regulatory oversight or supervision. This would promote business and commercial innovation, creativity, and entrepreneurship while enabling the government to operate more efficiently in meeting its goals.
The unprecedented opportunity of AI
Mindful that appropriate guardrails, governance, and protections need to be in place, today’s AI systems offer an unprecedented moment to augment regulators’ efforts in a way that makes people’s lives better. It is essential to determine how regulators can use AI safely and responsibly to improve regulation and its implementation. The U.S. government cannot afford to miss this opportunity, regardless of how independent agencies are characterized under the Constitution.
The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.
