Staff Reports
National Economic Council Director Kevin Hassett dropped a line on television that ought to make anyone who pays taxes and uses the internet sit up: the White House is “studying, possibly an executive order” to put powerful artificial intelligence models through an FDA‑style safety review before they’re released. That study moved fast — President Donald Trump signed Executive Order 14409, which sets up a voluntary pre‑release review process and tasks federal agencies with testing and cybersecurity coordination for frontier models.
What the White House actually put on the table
The executive order, while careful to call cooperation “voluntary,” lays out an administration‑led roadmap: agencies like Commerce, Treasury, NIST and the cyber shop will build processes for vulnerability assessments and a voluntary pre‑release testing path for “covered frontier models.” Hassett’s FDA analogy made headlines — “released to the wild after they’ve been proven safe, just like an FDA drug” — but the final text trimmed the heavy‑handed language from earlier drafts and left implementation to the agencies. The move was pushed by alarming demos of new model capabilities and warnings from industry researchers that some frontier systems could create cybersecurity risks if misused.
FDA‑style oversight for software? Don’t be naive
Here’s the problem: drugs and software are different animals. An FDA trial is slow, deliberate, and built for one‑time approvals; AI models evolve hourly and ship in updates. Any “pre‑release” friction that’s too long will simply reward the biggest firms who can swallow delays and costs — and squeeze out hungry startups and independent developers who actually drive useful competition. Picture a small team in Cleveland with ten engineers and a promising model; a drawn‑out, vaguely defined federal review could be a shutdown notice, not a safety net.
Real people will feel the tradeoffs
This isn’t just a tech‑nerd fight. Weak guardrails mean more cybersecurity holes for bad actors to exploit — banks, hospitals, and power grids aren’t immune to model vulnerabilities. Strong, slow rules risk handing the AI crown to whoever can shoulder the bureaucracy or to foreign rivals; as President Trump said bluntly, “whoever wins with AI wins,” and the administration refuses to cede that edge to China. So Americans face a plain choice: do we accept a bit of government friction now to reduce immediate risk, or do we gamble on speed and market forces and hope the market doesn’t burn down the barn?
There’s political theater, sure — Speaker Mike Johnson and other GOP leaders admit guardrails are needed while insisting competitiveness come first — but theatre won’t build a workable system. The executive order is a start, not a finish: it could either produce sensible, targeted security checks or become a paper shield that protects incumbents and leaves ordinary citizens exposed. Which will it be — protection that preserves American innovation, or protection that protects only a few powerful firms while everyone else pays the bill?
