For a moment in mid-September 2026, the biggest names in artificial intelligence sounded like they were on the same page about regulation. It did not last long. As The Verge put it, the AI regulation smackdown isn’t over: the industry’s rare show of unity on slowing down quickly ran into a White House that wants to keep accelerating.
The plan. On September 12, 2026, Anthropic CEO Dario Amodei published an essay titled “We Must Pace the Frontier”. Its central line is blunt: “We must slow the pace at which we improve the capabilities of AI models.” Amodei is not calling for a pause. Instead he lays out three steps:
- Embedded evaluators. Frontier labs would give third-party organizations such as METR “ongoing, employee-like access”, down to company badges, desks and laptops, so they can verify safety commitments and make sure incidents get reported. Anthropic says it is “unilaterally” committing to this step.
- Industry coordination. Leading AI companies in democratic countries would agree on common safety standards and limits on the rate of AI progress, with the US government helping to mediate those conversations.
- International agreements. The US and other democracies would attempt to coordinate with authoritarian governments on restrictions.
Amodei’s pitch to the skeptics is that pacing costs little: “Slowing the rate from ‘extremely fast’ to ‘only somewhat fast’ gives up relatively little strategic advantage.”
The unusual chorus. Rivals responded within hours. OpenAI CEO Sam Altman wrote, “I agree with Dario that we need to pace the frontier,” adding that the topic had been a primary subject of discussion inside OpenAI, and said his company would take the evaluator step too. Elon Musk needed three words: “Dario is right.” Google DeepMind’s Demis Hassabis was more measured: “Dario’s essay points towards the right path forward. The details need working through, but the direction is correct for meeting this critical moment.” He linked it to DeepMind’s own proposal for an industry-wide frontier AI standards body.
The pushback. Washington was far less enthusiastic. Speaking to reporters, President Donald Trump framed the question entirely around China: “We’re leading China in AI. We’re the most sophisticated country in the world, and frankly, I want to keep it that way, because whoever wins AI wins.” White House AI czar David Sacks argued that if the CEOs want to slow down, they can do it themselves without government involvement, and warned against creating a “duopoly” that freezes out competitors. House Speaker Mike Johnson sidestepped the question of legislation and suggested a meeting with “platform providers” instead.
Criticism also came from the other direction. Writing in Tech Policy Press on September 17, 2026, Dave Karpf argued that the embedded evaluator model risks resembling “Arthur Andersen’s relationship to Enron” and that a slowdown proposed by a company that is currently winning the race would mostly lock in the current standings. His conclusion: pacing the frontier should be a job for government enforcement, not for the labs themselves.
Why it matters. The episode shows where the fault lines now run. The largest labs broadly accept that they are moving too fast, yet the practical obstacles remain: fierce commercial competition, US-China rivalry, an administration that favors light-touch rules, and open questions about how independent in-house auditors can really be. Agreement on a direction, as Hassabis noted, is not the same as agreement on the details.
