There’s an old joke about State Department guidance. The President asks for a menu of options to deal with a crisis. State offers a three-checkbox memo. The United States may (1) wage global thermonuclear war, (2) do nothing, or (3) adopt State’s preferred policy position—which, whatever its downsides, doesn’t involve ending organized human life on the planet or letting the problem fester.

Something like that is going on with frontier-AI regulation. Presented with choices between letting free markets work and various forms of direct regulation (seizure, nationalization, moratoria), policymakers and industry executives are grasping for some kind of middle option between laissez faire and top-down control. One model that’s quickly gaining assent from national figures like Google DeepMind executive Demis Hassabis, swing-state Senator Dave McCormick, and Treasury Secretary Scott Bessent is a “FINRA for frontier-AI.”

Let’s quickly define both nouns. For the uninitiated, FINRA—the Financial Industry Regulatory Authority—is a so-called self-regulating organization (SRO) overseeing securities brokers throughout the United States. While FINRA has the functional equivalent of administrative law judges, reports to the Securities and Exchange Commission, and has regulations that may be approved, redlined, or vetoed by the SEC, the agency is formally organized as a § 501(c)(3) nonprofit corporation domiciled in Delaware.

The description NVIDIA offers for frontier AI is as good as any: “Frontier models are the most advanced AI models available at a given moment, trained on massive datasets to deliver state-of-the-art performance across many tasks, representing the leading edge of AI capability.” As models have grown more advanced and capable of performing complicated, agentic tasks, there’s an understandable concern that this bleeding-edge tech may carry out its programming in unpredictable ways or raise national security risks.

The basic gist of the “FINRA for AI” pitch is that the major firms can submit frontier models to a standards-setting body for both pre-release examination and post-release monitoring. As Mark Thomas, a proponent of the FINRA model aptly put it: “The key features are mandatory membership, industry funding, and a supervising [government] agency with authority to direct or veto the SRO’s actions. SRO rules bind members with the force of law, and the SRO enforces them directly, subject to appeal to the supervisor and then to federal court.”

There’s plenty of policy problems to unpack about this proposal. But there’s a more pressing objection: FINRA itself is unconstitutional.

A few weeks ago, WLF filed an amicus brief in Kielczewski v. SEC, where we argued that FINRA violates the Constitution’s Appointments Clause. America’s founding charter abhors unaccountable power. So in designing the Executive Branch, the Founders rejected proposals for distributing executive power to co-equal tribunes. Instead, they stood up a single, unitary President of the United States to faithfully carry out federal law. But as Fisher Ames noted in a 1789 House debate, the Executive Branch’s remit and authority is too much for “the circumscribed powers of human nature in one man” to cover. The President needs help.

But if Congress made the President’s helpers dependent on Congress or unremovable by the Chief Executive, assistance could easily become insubordination, undoing faithful execution. On the other hand, if the President could name all his own subordinates at whim, a faithless or incompetent White House occupant could run the government through officers of similar ineptitude or bad faith—also undermining faithful execution.

The Appointments Clause squares this circle: presidents can remove their subordinates at will, but those officers may acquire their commissions only by navigating a process that involves Congress, such as Senate confirmation. True, some officers may be appointed by the President himself—but only “inferior” ones with less direct responsibility. And even then, the President can appoint those officers only when Congress expressly lets him do so.

Real executive authority can’t be wielded outside this constitutional structure. That’s the problem with FINRA. It’s a serious regulator of the securities industry, enforces federal law, and functions as a sidekick to the SEC itself. Yet FINRA has no properly appointed officers conducting examinations, enforcing the securities laws, or writing regulations. (FINRA also denies its regulated community the full protections afforded by the Bill of Rights—yet another problem with this halfway house approach to applying state power.)

Perhaps the best outcome for frontier-AI regulation would be some compromise between government and enterprise (although we at WLF doubt it). But our Constitution is designed to take some choices, even ones that may be salutary, off the table. There can’t be a FINRA for AI for the same reason there can’t be a FINRA in the first place: the presidency can’t be privatized.