Regulation
A FINRA-style regulator for frontier AI has gone from a single Lawfare essay to a Treasury-developed proposal under White House review in four months, with Google, Anthropic and OpenAI already positioning for it
Mark Thomas's April proposal for a federally-supervised self-regulatory organisation modelled on FINRA has picked up backing from Google's own "FARO" framework, a public essay from DeepMind CEO Demis Hassabis, and now a Treasury Secretary Scott Bessent plan reportedly under active White House review.
Four months is a fast turnaround for a regulatory design to go from op-ed to policy pipeline, and the speed says something about how badly both sides want a settled answer. Thomas's core pitch borrows FINRA's structure directly: frontier developers above a compute threshold join a supervised body that writes and enforces rules, with a federal agency retaining approval power and the ability to write rules itself if the industry body stalls. That hybrid is designed to capture the thing pure government regulation struggles with, which is that the people who best understand frontier model risk are the labs building the models, and the thing pure industry self-regulation struggles with, which is that an unsupervised body's stable equilibrium is writing the minimum ruleset that satisfies the law and then stopping.
The structural details matter more than they look. An independent-majority board (12 to 15 seats versus 10 industry seats), collectively-elected rather than per-firm industry representation, and a dual-approval nomination process are all specifically aimed at preventing the outcome SROs are famous for, which is the regulated writing the regulations. Whether that holds in practice depends on whether the supervisor housed under Commerce actually uses its rule-writing threat credibly, because an SRO that never fears losing control of its own rulebook has no reason to write anything stronger than the floor.
What is notable is who is asking for this rather than resisting it. Google branded its own governance framework "FARO" in June, and Hassabis published an essay in July calling for exactly this kind of body and told press he wants it operational before year end. That is not the posture of an industry bracing for an unwanted regulator; it is the posture of an industry that has concluded a predictable supervised structure beats the alternative, which is ad hoc restrictions negotiated informally case by case with no durable rulebook behind them. A body they helped design is also, not coincidentally, a body whose rules they understand better than anyone examining them.
None of this closes the gap that actually touches trading platforms directly. An SRO governing frontier model developers regulates the labs building Claude, GPT and Gemini; it does not answer whether a broker-dealer using one of those models to generate investment guidance is doing something FINRA already has rules for, which is precisely the ambiguity a Robinhood lawyer flagged publicly this week. A FINRA for frontier AI could stand up fully and that specific seam, model-generated financial advice inside a regulated brokerage, would still be exactly as unresolved as it is today.
Read the original: Lawfare - A FINRA-style regulator for frontier AI has gone from a single Lawfare essay to a Treasury-developed proposal under White House review in four months, with Google, Anthropic and OpenAI already positioning for it. Commentary is the independent editorial view of Share Trading; the original article is credited to its publisher.