Staff Reports
Anthropic CEO Dario Amodei dropped a new policy essay this week titled “We Must Pace the Frontier.” It asks the industry — and asks Washington — to slow the pace of AI capability gains. The ask sounds noble until you look closer and see it could hand a regulatory moat to the very firms that built the tech. That’s not safety; that’s market protection dressed up as morality.
Amodei’s pitch: “We must slow the pace” — and he’ll police himself (with help)
Amodei is clear: “We must slow the pace at which we improve the capabilities of AI models.” Anthropic even pledged an operational step — to give third‑party evaluators employee‑level access inside the company to verify safety. The essay follows a viral resignation from a front‑line researcher who warned the industry was “gambling with our lives,” and an alignment lead who publicly estimated a non‑trivial chance of catastrophic risk within a decade. Those dramatic moments justify serious debate. But Amodei’s prescription is public and prescriptive: labs should pace capability jumps and governments should help set the rules.
Why the plan risks becoming regulatory capture, not safety
Here’s the rub: when frontier labs ask regulators to backstop their strategy, they invite a permanent winner‑take‑all regime. Requiring expensive pre‑deployment tests, embedded evaluators, and heavy compliance will favor deep‑pocketed incumbents and block startups. That’s not an accident. It’s a predictable outcome of well‑meaning safety rules drafted by people who already sit at the top. And yes, worrying about China’s theft of cutting‑edge models complicates the picture — but handing the US government the power to effectively freeze the frontier in place risks ossifying American champions just as adversaries sprint another direction.
Politics, policy, and what conservatives should demand
Lawmakers are rightly circling. Speaker Mike Johnson and President Trump have publicly questioned calls for locking in lab control, and members of both parties are pushing for hearings. Conservatives should support oversight — not a cozy deal that hands industry insiders a regulatory shield. Demand transparency, insist labs show they can and will self‑slow when needed, and resist one‑size‑fits‑all mandates that bake in market advantage. If safety rules are created, write them to be time‑limited, contestable, and designed to encourage competition — not to protect elites who want a permanent seat at the table.
In short: Amodei and Sam Altman can and should slow down their own work if they truly fear doom. If they want government help, fine — but don’t let “safety” be the Trojan horse for entrenching monopoly power. Competition, clear oversight, and clever export controls are better tools than a permanent regulator‑approved club for Big AI. Call it tough love: worry about extinction, sure — but don’t let fear be used as cover for corporate cartels. Washington would do well to listen, but not to sign a blank check.
