In "Is Private AI Regulation Constitutional?" (July 10), Andy Jung questions the constitutionality of the frontier-AI self-regulatory organization I described in Lawfare in April and that Google recommends in its recent policy paper. He is correct that a government agency with real authority must supervise such a body. The reasons are both legal and practical, because a weak supervisor permits the industry to avoid regulation altogether. But recent court cases do not cast this design in doubt so much as clarify what it requires.
The Fifth Circuit's repeated objection to the horseracing authority is that its supervising agency had less control than the Securities and Exchange Commission (SEC) holds over the Financial Industry Regulatory Authority (Finra)—the model for my original proposal. The Eighth Circuit, which Mr. Jung does not mention, upheld the horseracing authority because Congress modeled it on the securities scheme. Even the courts most skeptical of private regulation treat Finra as the benchmark.
The D.C. Circuit did find a defect in Finra's procedures in Alpine Securities, on a preliminary record. It held that Finra could not expel a member, shuttering its business, before SEC review. Lesser, reversible sanctions, the court noted, raised no such problem. Finra amended its rules to require SEC review before expulsion.
These precedents don't make the Finra model unworkable for AI, they tell Congress how to draft the enabling statute. And the legal requirements point the same direction good design does: a supervising agency with genuine control over rules and sanctions, and an industry body that remains nimbler, more expert and more innovative than direct regulation could be.