I think you’re exactly right that there’s no “FDA for AI” coming, and that real power is sitting in procurement, standards, and sectoral regulators rather than in some future super agency.

However, I believe that people calling for an “FDA for AI” aren’t actually begging for another bloated, bumbling bureaucracy staffed by Silicon Valley’s benchwarmers. What they are saying is that they don’t trust pure techno optimism to safeguard their communities. They understand that when founders talk about “moving fast and breaking things,” the “things” at risk include their jobs, their data, and their institutions. And when a lucky few stumble into a market-defining product, they will quite rationally turn that success into political influence over the people whose things they just broke. (sound familiar?)

Our job, as lawyers who live in the real world, is not to hate founders or VCs—it’s to meet them standing up, to insist on rules that protect the public, and to warn our clients when a clever technology play is on a collision course with the people it depends on.

But that’s exactly why we should be nervous about who is steering the ship.

If you want a historical analogy, AI today looks less like the Manhattan Project and more like what would have happened if we had outsourced the Manhattan Project to a VC‑backed startup in 1939 and told ourselves that “the market” would sort out nuclear safety.

With nuclear, the state eventually did something only a sovereign can do: it centralized talent and resources, built the bomb, used it, and then scrambled to build a regime around proliferation and control. We didn’t leave the question “who gets nuclear weapons?” to private labs optimizing for growth.

With AI, we’ve mostly done the opposite. We outsourced the Manhattan scale work to private firms, celebrated the demos, and only afterward started asking basic sovereign questions: who should have access to this capability, for what purposes, and subject to which non negotiable constraints?

The standard reassurance is that “the market will handle it.” But that’s a cartoon version of Adam Smith. Smith took for granted law, courts, property, and public goods; he worried about collusion and power. Markets, in his view, are engines for allocation, not systems for deciding how far to push dual use technologies that can destabilize security, finance, or information environments.

So I’m with you that we don’t need an “AI FDA,” and I share the concern about building another slow, captured bureaucracy. But I don’t think we can treat contracts and industry led standards as sufficient either. For nuclear, biohazards, controlled chemicals, and certain weapons, we layer licensing, export controls, liability, criminal penalties, and public education on top of markets.

We accept that a sovereign—with courts and, if necessary, people with guns behind it—draws the red lines.

AI belongs in that family. Not because it’s magic, but because it’s general purpose, cheap to copy once built, and easy to couple to critical systems.

In other words: yes, governance is embedded. But if we don’t make the sovereign intent explicit—what must never be done with these systems, and under whose authority—then we haven’t really moved beyond hoping that the Manhattan Project being run out of a cap table somehow converges with the public interest.

Thomas View, Esq.

TEMVI Legally Assured Cybersecurity