U.S. AI labs want to “pace the frontier,” citing safety fears. The fine print tells a different story.
Anthropic and OpenAI are pushing for independent inspectors and government rules to slow down the pace of AI development. If that framework raises the cost of entry and innovation, “pacing the frontier” starts to look like pacing the competition.
BECOME A MEMBER
Article by Rebel News staff.
Tonight on The Ezra Levant Show: the leading AI laboratories are calling for the frontier to be paced, or at least for their competitors to be, even as the structure of their proposal would raise the cost of entry for everyone else.
Dario Amodei, chief executive of Anthropic, published "We Must Pace the Frontier." Sam Altman of OpenAI said he agreed. Neither executive announced a pause at his own company. Both pointed toward outside inspectors, industry standards and government coordination.
Amodei writes that AI systems are improving in ways that outrun human engineers, and cites incidents in which swarms of AI agents conducted unauthorized cyberattacks. He warns that within months such a swarm could threaten the internet at a scale measured in hundreds of billions of dollars. His remedy is institutional: permanent evaluators inside the major labs with employee-like access; common standards among democratic-country labs, with Washington brokering an antitrust waiver so competitors can coordinate; and some form of understanding with China.
Altman endorsed pacing the frontier and said OpenAI would admit independent evaluators. The wider framework of standards, checkpoints and official oversight remains Amodei's. When the two dominant firms jointly ask the state to write the rules, the accountability question is whose interests those rules will serve.
Dario has written that we need to “pace the frontier,” and Sam has agreed. People may be surprised by my response: go ahead.
— David Sacks (@DavidSacks) September 13, 2026
You guys are the frontier. By any reasonable metric — market share, revenue growth, model capability — the two of you have a duopoly on frontier…
David Sacks, co-chair of President Trump's Council of Advisors on Science and Technology, answered directly. By market share, revenue growth and model capability, he argued, Anthropic and OpenAI already function as a duopoly. "Go ahead. You guys are the frontier," he wrote. The simplest way not to build superintelligence is for those companies to refrain from building it. Demanding a preferred regulatory framework as the condition of that restraint, he said, would look like blackmail; refusal to slow down without that bargain would reveal another attempt at regulatory capture.
If the risk Amodei describes is genuine, Anthropic can reduce its own contribution without waiting for rivals or Beijing. Permanent inspectors are a cost startups cannot easily bear and incumbents can. Release checkpoints operate like licences that favour firms already staffed to navigate them. A pace negotiated under an antitrust waiver resembles a cartel described as safety policy.
China exposes the verification gap: if democracies slow and Beijing defects, Amodei notes, China could hold the most powerful technology in history. An advanced model is a file that can be stolen, copied and imitated. Unlike warheads countable from space, a file cannot be verified the same way. "Pace the frontier" then becomes an honour system that binds only the honourable, plus a permanent government role of keeping everyone else behind.
Allum Bokhari has argued that veterans of the old "trust and safety" offices have moved into AI safety roles, where "safety" often means "machine learning fairness": constraining models so they do not register differences between groups. When Amodei seeks inspectors with power to publish findings inside every major lab, the public is entitled to ask whether those inspectors will audit for rogue systems, or decide which claims are allowed online.
Sacks has previously accused Anthropic of regulatory capture through fear and of advancing "woke AI" through blue-state legislation. An industry letter this summer endorsed open-weight models anyone can download and run. Meta signed; Anthropic did not. Amodei has also urged a crackdown on industrial-scale distillation, how smaller labs close the gap by querying larger systems.
Anthropic's unilateral inspector program remains most under its control. The steps that would bind it require Washington and Beijing, and therefore time. Bernie Sanders has announced a permanent ban on "superintelligence" and a pause on advanced development. The slowdown coalition stretches from a democratic socialist to two of the industry's richest executives. They do not share identical statutes, but they share a direction: slow the work and enlarge the state's role. The only remaining disagreement is who sits on the commission.
If advanced AI is dangerous, liability is a transparent rule. If it is powerful, dispersing it among more builders is a different model than concentrating it in two laboratories with government-approved inspectors. The laboratories say the machines must slow down. The process they are designing would also slow everyone who might build something outside their control.
GUEST: John Carpay, President of JCCF on his new report regarding the Notwithstanding clause.
COMMENTS
-
Bruce Atchison commented 2026-09-14 20:48:45 -0400 FlagRemember that socialism runs communication avenues while communism owns them. So throttling AI is a back door for socialists to control what we consume. Leftist governments can’t survive without censorship because their philosophy is bankrupt. People naturally crave freedom but socialist and communist governments forbid exposing their faults.
-
Jane Vandervliet commented 2026-09-14 20:45:00 -0400 FlagThe Libtards and the Conswervatives are of exactly the same worldview…..woketardism.
