In brief...
At a conference co-hosted by the Regulatory Studies Center and Norm Ai, speakers discussed challenges and opportunities for artificial intelligence to streamline regulatory compliance burdens and improve stakeholder experiences with government services. This commentary summarizes introductory remarks by CFTC Commissioner Kristin Johnson and Ohio State University Law professor Bridget C.E. Dooling.
The Regulatory Studies Center’s July 8 conference, “Can AI Streamline Regulation and Reduce Compliance Burdens?”, co-sponsored by Norm Ai, opened with a keynote session that reflected one of the core tensions raised by moderator Roger Nober, Director of the Center: Can artificial intelligence (AI) make government more efficient without compromising the accountability that public institutions are designed to uphold?
The session featured a keynote address by Commissioner Kristin N. Johnson of the U.S. Commodity Futures Trading Commission (CFTC), who joined virtually, in-person remarks by Professor Bridget C.E. Dooling of The Ohio State University Moritz College of Law and former Research Professor at the Regulatory Studies Center, and introductory framing and moderation from Director Nober. Together, they offered not just a roadmap for AI’s adoption in regulatory agencies but a meditation on the balance between innovation and institutional responsibility.
Setting the Stage: A Slow Machine Meets a Fast One in the Age of AI
Director Nober opened with a framing that grounded the day in realism. While much of the excitement around AI is focused on speed, prediction, and automation, he reminded the audience that regulatory bodies are intentionally built to move slowly in order to consider, consult, deliberate, and, when necessary, coerce. This isn’t inefficiency, he argued, but a feature of constitutional design. “Adopting new technology is one thing,” he noted, “but adopting it in a regulatory framework is something else.” In the public sector, where the stakes for error are high and the space for redress is narrow, caution is not optional—it is foundational.
Drawing from his previous government experience, Nober illustrated this tension through a series of analogies. After 9/11, aviation authorities borrowed fraud detection models from credit card companies to more expeditiously screen passengers. The lesson he suggested was clear: the government can borrow speed, but it cannot outsource accountability—a principle that underscores the distinct obligation of public institutions to remain answerable for decisions made in the public’s name, regardless of the tools employed. He reinforced the point with a familiar image: though modern aircraft can fly and land autonomously, no one boards a plane without a human pilot. In regulation, as in aviation, human judgment and public trust still matter. As agencies explore AI integration, Nober left the audience with a pointed question—one that would echo throughout the session: What remains uniquely human in a system increasingly shaped by machines?
Commissioner Kristin Johnson: Supervision, AI-Washing, and Interagency Collaboration
Following Director Nober’s introduction, Commissioner Johnson opened with an expansive view of the evolving relationship between financial regulation and AI. She situated her remarks in a global context, referencing a recent International Monetary Fund study cataloging the accelerating adoption of AI in finance, from customer service chatbots and trading algorithms to anomaly detection in market surveillance designed to detect fraud and market manipulation, as well as cyber defense systems. While many discussions on AI tend toward the abstract, Johnson’s remarks were highly concrete. She outlined how AI is already operationalized in critical domains such as anti-money laundering, internal fraud detection, and sanctions compliance. Just as importantly, she described how regulators, including the CFTC, are beginning to respond. The agency’s recent Request for Comments on the Use of Artificial Intelligence in CFTC-Regulated Markets and its December 2024 Staff Advisory on Use of Artificial Intelligence signaled more than just fact-finding; the Commission staff publicly affirmed that the use of AI does not displace legal accountability.
One of the more sobering moments came with her warning about “AI-washing,” the practice of overstating or fabricating AI capabilities to attract investment or mislead regulators. She cited a landmark enforcement action in which a company falsely marketed a proprietary AI trading bot that did not exist, an act of deception that led to a $1.7 billion penalty, one of the largest in the Commission’s history. The case illustrated how AI’s mystique can become a smokescreen for fraud, underscoring the need for greater technical fluency among enforcement teams.
To meet these emerging threats, Johnson outlined two specific policy proposals. First, she called for heightened civil monetary penalties for bad actors who intentionally misuse AI to commit fraud, evade regulatory oversight, or mislead vulnerable investors. The deterrent value of such penalties, she argued, must match the scale and sophistication of the tools being abused. Second, she advocated for the creation of an interagency task force, made up of the nation’s financial regulators, to collaboratively design and implement guidelines and frameworks for AI governance across financial markets.
Johnson’s remarks offered a balanced perspective: recognizing AI’s inevitability, acknowledging its risks, and calling on regulators to evolve not only their tools, but also their institutional posture. If AI is to strengthen rather than destabilize market integrity, she argued, the regulatory community must lead and stay ahead of the risks.
Professor Bridget Dooling: Beyond “Human in the Loop” – Preserving Judgment and Reasoning
Prof. Bridget Dooling followed with an in-person keynote that challenged one of the day’s most frequently used terms: “human in the loop.” Rather than accepting it as a default safeguard, she argued that the phrase subordinates human judgment to post hoc approval of machine outputs and understates the important human role needed in many situations. Dooling’s critique was neither Luddite nor reactionary. On the contrary, she acknowledged the strengths of AI in domains like pattern recognition, route optimization, and communications surveillance. But she drew a hard line around decisional contexts, those moments in governance where human judgment is not perfunctory, but integral.
Her underlying concern wasn’t with automation per se, but with epistemic complacency: the belief that inserting a human checkpoint at the end of an AI workflow can substitute for genuine public reasoning. In this context, she emphasized a crucial distinction between editing and writing—between reviewing machine-generated outputs and engaging in the intellectual work of drafting, deliberating, and justifying a decision. In regulatory settings, she argued, this assumption is dangerous. Decisions, especially those that interpret law, weigh competing interests, or impose burdens, require real reasoning, not just review.
One of the most intellectually urgent portions of Dooling’s keynote came when she turned to generative AI and the drafting of regulations. In light of the fact that many agencies are experimenting with large language models (LLMs) to accelerate rulemaking, Dooling raised a critical warning: writing is not merely documentation; it is cognition. Drawing on her own experience in government and academia (including her paper “Ghostwriting the Government”), she argued that writing is how regulatory agencies think through problems, test assumptions, and structure obligations. Automating that process doesn’t just save time; it risks outsourcing the very act of public reasoning that underpins legitimacy under the Administrative Procedure Act. This wasn’t a nostalgic defense of pen-and-paper bureaucracy. Rather, Dooling was pointing to a deeper truth: when we automate expression, we may unintentionally hollow out deliberation. A rule that “sounds” correct, but bypasses real deliberation, is a brittle foundation for coercive state authority.
Q&A: When Should AI Stop at the Door?
The Q&A session extended the discussion into practical frontiers. In response to a question about automated traffic enforcement, Dooling noted that she is accepting of machine-imposed fines, especially for binary violations like speeding or stop-sign violations. But she asked the audience to consider what is lost when government actions carry no human signature or recourse. Similarly, when asked about AI-assisted tax audits, she suggested that while automation may be unobjectionable for simple returns, it becomes more problematic as complexity and discretion are introduced.
In response to public comment analysis, Dooling acknowledged that agencies already utilize de-duplication tools to manage large volumes of similar comments—a practical necessity in the digital era. But she cautioned against expanding AI’s role to include summarizing or triaging. Agencies are not simply data processors, she reminded the audience; they are institutions of democratic accountability. Letting a language model decide what counts as “representative” or “important” risks flattening pluralism and missing dissent.
Finally, on the provocative question of using AI to eliminate outdated regulations, Dooling drew a line. While analytics can help identify potential friction points, deregulation, like regulation, requires reasoned justification. You cannot delegate the procedural steps for repeal to AI or LLMs without eroding the legitimacy of the rulemaking process itself.
Final Thoughts: Guardrails Before Gains
This opening session of the conference offered more than an update on agency tools or a forecast of regulatory trends. It was a sober exploration of what it means to govern in the age of artificial intelligence. Commissioner Johnson called on agencies to modernize, not just technologically, but structurally and ethically. And Professor Dooling reminded us that governance is not a technical problem to be solved, but a human endeavor to be preserved.
If AI is to truly streamline regulation rather than obscure it, the burden is not on the machine; it’s on us. And that burden includes thinking clearly, reasoning thoroughly, and always asking not just what AI can do, but whether it should be allowed to do it in the name of the public interest.