AI regulatory framework is missing from proposals, Greifeld says

Bob Greifeld, managing director and co-founder at Cornerstone Financial Technology, said the U.S. debate over an AI regulatory framework is producing thousands of measures without a comprehensive, long-term structure. In a CNBC opinion piece, the former Nasdaq executive argued that state legislatures, Congress, the executive branch and private market participants are addressing fragments of the issue while artificial intelligence capabilities continue to advance.

Greifeld wrote that states have more than 1,500 AI-related bills under consideration, while Congress has hundreds of proposals and the executive branch has issued dozens of actions. He said the volume of activity, approaching 2,000 proposals, shows that market forces alone are unlikely to govern the technology adequately.

His central claim is that many proposals may be useful in the near term but do not create a durable national system for overseeing AI as its uses and risks evolve. Greifeld called for policymakers to create a national regulatory body with a broad mandate to supervise the technology across current and future applications.

What would an AI regulatory framework do?

An AI regulatory framework would set rules, review mechanisms and institutional responsibility for how artificial intelligence systems are developed and deployed. Greifeld’s argument is that a single national body could adapt oversight as the technology changes, rather than relying on separate bills and agency actions aimed at today’s concerns.

Greifeld said regulators often arrive after a crisis, citing the Securities and Exchange Commission’s creation after the 1929 market crash as an example of a national body built in response to systemic failure. He argued that policymakers should not wait for a comparable shock in artificial intelligence before establishing permanent oversight.

He acknowledged that a national AI regulator would not solve every problem. Greifeld said such a body would need scrutiny from Congress, the executive branch, the courts and the public, and he warned that regulators can overreach. He also described AI as part of a global race, requiring policymakers to balance innovation with public safeguards.

Why Greifeld points to the SEC model

Greifeld drew on his experience at Nasdaq to explain one mechanism he believes could inform AI oversight. He wrote that changes to core exchange technology had to be submitted to the SEC, which would publish the details for public comment and later review.

That process meant competitors could see planned changes before implementation, a feature Greifeld said was uncomfortable from a business perspective. Over time, however, he argued that this disclosure-and-comment system helped support the strength of U.S. capital markets.

Greifeld suggested that major changes to large language models could be subjected to a comparable public review process, though he said the response would likely be far larger than the comment flow generated by exchange order-type changes. A large language model is an AI system trained on extensive text and other data to generate or process language, and changes to such systems can affect users, developers and businesses that rely on them.

Greifeld said he was uncertain whether a national AI regulator should be viewed as the greater good or the lesser evil. His conclusion was that regulation could slow progress in the short term, while clearer rules could create better conditions over the long run.

This story draws on original reporting from CNBC.