Sam Altman used a Monday post on X to name, for the first time in this much detail, the two outcomes he fears most from AI development: humanity “losing control of the future to AI,” and power concentrating around one person or company. The post landed two days after Altman and Elon Musk both publicly backed an essay from Anthropic chief executive Dario Amodei urging frontier labs to slow the pace at which they improve their most capable models.
That is the news here, not the essay itself. Amodei’s call for a deliberate pace already carried weight on its own. What changes the story is that OpenAI’s chief executive and Elon Musk both put their names behind it within the same weekend, and that Altman went further by sketching the specific failure modes he is trying to head off.
“We welcome a federal framework that sets consistent safety requirements for frontier AI,” Altman wrote, adding that “no amount of American competitive pressure should justify recklessness.” He also wrote that “committing to having independent evaluators with employee-like access is a great idea, and we will do the same,” echoing a commitment Amodei says Anthropic is already making.
Altman was careful to draw a line between his position and an actual pause. “When we talk about ‘pacing,’ we do not mean ‘stopping,’” he wrote. “Progress has been rapid and will continue to be.” That distinction matters more than the show of unity: three rival chief executives agreeing in public that AI needs a speed limit is not the same thing as any of their companies changing a release schedule, and neither CNBC’s report nor the underlying posts describe one.
Musk’s contribution, as CNBC characterized it, was to back Amodei’s essay alongside Altman over the weekend. The source does not attribute a direct quote to Musk on the matter, only the joint public support.
The timing compounds the pressure on all three companies. Amodei’s essay followed an Anthropic researcher’s resignation last week, in which the researcher warned that people building the technology privately believed it could “kill us all by the end of the decade,” a claim that reportedly pushed employees at both Anthropic and OpenAI to raise concerns about catastrophic risk. AI stocks fell Monday as investors weighed the CEOs’ comments, and President Donald Trump dismissed the warning the day before, arguing a slowdown would hand China the lead in AI.
Amodei’s own proposal, which he says Anthropic will follow regardless of what rivals do, rests on three planks: give outside evaluators “employee-like access” to frontier models, agree on common safety standards across labs, and coordinate the pace of development internationally where possible. He has acknowledged that the last piece may not survive contact with geopolitics. Speaking to CBS News, he called it “the toughest dilemma,” since any country that keeps racing ahead gains a military and commercial edge the others would be giving up. China’s Global Times and its Foreign Ministry both dismissed the CEOs’ comments Monday, the latter calling them “fearmongering.”
None of this changes what ships next. Anthropic and OpenAI are both moving toward what CNBC described as historic public offerings, and Altman has already ruled out an OpenAI IPO within 2026. A company preparing to go public has every incentive to be seen managing risk in public, whether or not it changes a single training run behind the scenes.
For operators building on frontier models, the signal to track is not this week’s statements but the next model release from any of the three labs. If a scheduled capability jump slips, or if an external evaluator’s access becomes contractual rather than promotional, that is the point at which “pacing” becomes a real constraint on the roadmap rather than a talking point.
Reporting by Kai Nicol-Schwarz for CNBC, published September 14, 2026.
