TechNewsFirst

Over the weekend of September 12 and 13, 2026, Anthropic co-founder Dario Amodei published an essay called "We Must Pace the Frontier," arguing that frontier AI labs should coordinate to slow capability gains and warning that AI agents could "take over the entire internet" within six to twelve months without new safety standards. Sam Altman, Demis Hassabis and Elon Musk all put their names to it. The same weekend, an Anthropic-linked safety researcher named Jacob Coxon publicly announced he was quitting the industry, saying he no longer believed labs, including his own, could control what they were building.

By Monday, three of the most powerful people in the country who are not running an AI lab had each rejected the premise, and no two of them rejected it for the same reason.

The short seller: there's nothing here to slow down

Michael Burry, the investor who bet against the 2007 housing market, posted his response to X on Sunday night, September 13. Mediaite ran the story the next day. Burry's argument has four parts. Large language models, he says, are not AI and will not become artificial general intelligence, so there is no runaway capability to pace. Rising competition from rivals threatens the incumbents putting their names on safety pledges, and a coordinated slowdown protects market position more than it protects anyone from Skynet. IPOs need hype, and "we built something so powerful it could be dangerous" is a marketing line dressed up as a warning. Put together, Burry's read is that the safety announcement is cover for something duller and more damaging to the labs: growth that is actually slowing, for ordinary technical reasons, right as IPO paperwork comes due.

That's a market skeptic's argument. It says the danger is fake because the underlying technology has hit a ceiling.

The president: there's nothing here, period

Donald Trump didn't engage with any of that. In Ireland on Sunday, September 13, he dismissed the safety warnings as exaggerated and blamed unnamed "negative forces." By Monday, September 14, he'd moved to Truth Social. "AI taking over the World, destroying Humanity...is a HOAX," he wrote, adding that "the only control or 'guardrails' that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT" and that "there is a SICK conspiracy going on against AI and Data Centers." NBC News published the posts the same day. CNBC followed on September 15 with a piece on how squarely Trump's position cuts against the industry's own stated request for a slowdown.

Trump's version isn't a technical argument at all. He isn't saying the models can't do what Amodei says they might do. He's treating the entire safety conversation as an attack on American AI and data center buildout, full stop, the same register he uses for other policy fights he frames as rigged against American industry.

The vice president: the danger might be real, the fix is the problem

JD Vance took a third position, and it doesn't fit neatly next to the other two. On the All-In podcast on September 15, responding directly to Amodei's essay, Vance asked why "the people who are at the frontier of the AI economy are throwing up their hands and saying, well, we've built Frankenstein, and the solution to Frankenstein, apparently, is to create a one-world governance structure for artificial intelligence." His line to the labs: "If you're building Frankenstein, stop." He called the industry's turn toward government regulators "a bit of a Trojan horse," reasoning that incumbents asking Washington to regulate them is a good way to make sure nobody smaller ever catches up. HuffPost carried the exchange the same week.

Here's the part that gets lost if you lump Vance in with Trump: by most reporting on the podcast, Vance did not call Amodei a liar or a fraud. He said he believes Amodei is sincere about the risks he describes. What Vance rejects isn't the concern, it's the proposed cure, a global governance structure he treats as a decision better left to engineers than bureaucrats, and one that happens to lock in whoever gets there first. That's a regulatory-capture argument, not a doom-denial argument, and it's a genuinely different claim from either Burry's or Trump's.

This isn't new for Vance. At the Paris AI Action Summit in February 2025, he told the room he wasn't there to talk about "AI safety," he was there to talk about "AI opportunity," and used the speech to criticize the European Union's AI rules directly. TechCrunch covered it at the time. September 2026 is the same instinct, aimed at American labs instead of European regulators.

Four different objections, one outcome

Lay the four positions side by side: Burry says there's no capability worth pacing. Trump says there's no danger worth discussing. Vance says the danger may be real but the proposed fix hands power to the wrong people. And this outlet's own prior reporting argued a fourth thing entirely: that the labs' synchronized "we're slowing down for safety" announcement landed the same weekend their IPOs and post-lockup share sales were coming into view, and that timing is worth noticing on its own, regardless of whether the underlying models are actually dangerous.

None of these four arguments agree with each other on the facts. A market skeptic who thinks LLMs have plateaued and a sitting vice president who takes Dario Amodei at his word about the risk are not making the same case, even though both of them end up opposed to the same policy outcome: no coordinated slowdown, no new global governance body, business as usual for the labs racing each other. That convergence is real. It is also not evidence that any one of the four is right. Four people can agree on what not to do for four incompatible reasons, and all four can still be wrong.

What to actually watch

The scoreable claim in all of this isn't "AI is dangerous" or "AI is a hoax," both of which are close to unfalsifiable on any near-term horizon. It's Amodei's specific six-to-twelve-month window for AI agents "taking over the entire internet." That has a clock on it. Mark September 2026 as the start date, and revisit it in the spring and again by next September. If nothing resembling that happens, Burry's read looks better. If something does, the "hoax" and "Trojan horse" framings age very badly, very fast.