Anthropic official brand assets (anthropic.com)
FTC chair warns the proposal could insulate incumbents while rival CEOs distance themselves from the ask
Dario Amodei wants the biggest AI companies to sit in a room and hash out safety standards together. Washington’s antitrust cops would rather they didn’t.
The Anthropic CEO published an essay on September 12 outlining what he calls “pacing the frontier,” a framework for keeping AI safety efforts in lockstep with the breakneck pace of capability development. The centerpiece: a request for a narrow antitrust waiver from the US government that would let competing AI firms coordinate on safety protocols without triggering legal liability.
The proposal, and who’s not buying it #
Amodei’s essay laid out a three-part strategy. First, embed independent third-party evaluators inside AI labs to monitor safety and alignment work in real time. Second, establish coordinated safety standards among democratic nations. Third, open dialogue channels with authoritarian states, including China, to reduce the odds of a global AI arms race spiraling out of control.
The first two pillars drew broad, if cautious, support from the industry. The antitrust waiver did not.
FTC Chair Andrew Ferguson weighed in on September 15 with pointed skepticism. His concern is straightforward: granting antitrust exemptions to the most powerful AI companies could raise barriers for new entrants, effectively locking in the advantages of the current incumbents.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Industry peers keep their distance #
OpenAI committed on September 13 to the evaluator access model, pledging to let third-party monitors inside its operations. Sam Altman endorsed that piece of the framework. But he stopped well short of backing the antitrust waiver itself.
Nvidia CEO Jensen Huang took a cleaner line on September 15, stating that existing laws are sufficient to support the safety efforts the industry needs. No new antitrust carve-outs required.
Why this conversation is happening now #
The timing of Amodei’s essay isn’t coincidental. The AI industry is grappling with recent warnings about self-improvement capabilities in frontier models, the kind of recursive learning loops that could produce systems significantly more capable than their creators intended. Amodei’s essay explicitly references concerns about imminent risks from AI advancements projected within the next six to twelve months.
That timeline matters. If frontier models are approaching thresholds where they can meaningfully modify their own architectures or training processes, the window for establishing safety coordination narrows fast. Amodei’s argument is essentially that the speed of progress makes voluntary, company-by-company safety efforts insufficient.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our