Forty-eight hours after Dario Amodei's pacing-the-frontier proposal, the negotiation got swallowed by politics: seven Truth Social posts declaring AI doom a 'hoax,' a surprise Oval Office call to Jensen Huang live on stage, and Obama publicly backing the slowdown. Plus: Terence Tao leads 25 Fields medalists in declaring the labs 'severely misaligned' with mathematics itself.
The slowdown debate just went fully partisan.
Dario Amodei's pacing proposal was supposed to open a negotiation between the labs and society over who controls the speed of AI. Instead, Trump spent Monday declaring the whole thing a hoax across seven Truth Social posts, Obama publicly endorsed the slowdown, Kamala Harris piled on, and the safety community watched its nightmare scenario unfold in real time: once AI safety belongs to a political tribe, half the country will oppose it simply because the other half supports it. The next thing to watch is whether this jumps from politics to policy.
Yesterday's argument holds: the pacing-the-frontier proposal plus the other lab leaders explicitly agreeing — even reposting Dario's article — marked a meaningful inflection point, a recognition that the labs will no longer be the sole arbiters of how fast new AI comes to market. What's being negotiated now is how that power gets distributed.
Martin Casado, initially hostile to the letter, now argues it's entirely consistent to applaud the labs' attempt to find a governance approach, express skepticism about the specific proposal, and worry about the broader 'safety complex' — as long as you don't conflate the three. He still calls heavy-handed federal regulation 'disastrous' and likely without coordination, but concedes 'the labs really are trying to do the right thing here.'
The President's Monday barrage claimed the only guardrail AI needs is 'a strong and smart high IQ president,' accused Anthropic's Dario of 'pretending to be a perfect little angel,' warned of a 'sick conspiracy' against AI and data centers whose only happy party is China — and put 'conspiracy theorists, treasonists, traitors, and leakers' on notice.
Trump called AI and data centers 'the greatest economic development engine in history, bigger than oil, gold, diamonds, or even the Internet' — and complained that Google wants to build a massive plant in Finland because U.S. permitting is too difficult. Meanwhile, per Trump, President Xi just announced China will do 'absolutely nothing to stand in the way of AI or its future.'
— President Donald Trump, on Truth Social. Trump grouped AI doom with 'Russia, Russia, Russia,' both impeachment 'hoaxes,' and the 'global warming scam,' casting slowdown advocates as radical-left revolutionaries 'working for people that do not have the best interests of the United States in mind.' Later: 'I am the hoax buster.'
— Vice President JD Vance, on the labs' calls for regulation. Speaking at Andrews Air Force Base, Vance said he feels 'a little bit weird' about frontier AI companies 'begging the government to regulate them.' The Department of War's CTO account piled on with a graphic reading 'AI First War Department' and the copy: 'Americanism, not effective altruism.'
AI safety advocate Jeffrey Miller pointed to a stack of polls showing Americans are broadly uneasy about AI and warned that branding everyone concerned about safety a conspiracy theorist or traitor is imprudent for Republicans heading into the midterms. Either way, the lines are officially drawn: Trump thinks the extinction argument is a hoax and wants America building as fast as possible.
— Jensen Huang, NVIDIA CEO, at the All-In Summit. Asked at the All-In Summit how to explain a 10% chance of AI exterminating humanity to the public, Huang called such predictions troubling, alarming, and irresponsible — then ran through the doomers' track record: AI replacing radiologists, eliminating white-collar work, generating 90% of code. All wrong. 'We have to take accountability for all of the stupid predictions that were made.'
In a twist that would have been rejected from Hollywood writing rooms for being too unrealistic, Huang took a surprise Oval Office call live at the All-In Summit. Trump repeated the hoax framing on speakerphone — 'the robots are not going to be taking over the world' — and Jensen replied, 'We're not going to let it happen, sir.'
Beneath the showmanship, Huang supported embedded independent auditors at the labs (so long as they're truly independent of the doomer community) and framed safety as an engineering problem: root-cause incidents like Hugging Face, then institutionalize the fix. Danger concentrates at the frontier labs, he argued, because only they can hand agents access to near-infinite compute.
After privately urging Democrats to make AI a campaign platform — reportedly saying it would be 'one of my central agendas' if he were running in 2028 — Obama took his views public on X, calling the lab leaders' agreement to slow down 'a good and necessary first step.' For someone historically disinclined to wade into policy discourse, the post itself is the signal.
— Barack Obama, on X. Obama positioned himself as neither accelerationist nor doomer: whether AI delivers breakthroughs in medicine, energy, and education or unleashes huge disruption, inequality, and catastrophe 'will depend on the choices that we make now' — choices made not just by the companies involved, but by all of us.
Obama announced the initiative from Ron Conway's venture firm, whose stated goal is to 'rebuild bonds between leaders in the artificial intelligence sector and Democrats who are demanding new rules to regulate the technology.' Mike Solana's read: AI safety is now irreversibly polarizing, with an information environment at least as difficult as COVID's.
With Kamala Harris lending her weight to the slowdown calls, the tribal sorting is underway: Democrats line up behind slowing down, Trump declares the whole thing a hoax — and, as observers across the spectrum lamented, half the country will now oppose AI safety simply because the other half supports it.
The Nasdaq stayed roughly flat Monday, but the semiconductor index lost 5.9%, and financial media is floating rotations into SaaS — or even the much-maligned Indian IT stocks — as places to hide while the AI sector wobbles. Meanwhile, life goes on: IPO plans are moving forward and the big labs are racing to adapt to the new era.
Seven Truth Social posts and an Obama endorsement later, the discourse has fully polarized — but so far it's all rhetoric. The signal that matters now is whether the hoax-versus-slowdown fight starts producing actual legislation and rules.
Before the slowdown debate consumed everything, the biggest story in AI was OpenAI solving a Millennium Prize problem — with NYU's Tristan Buckmaster questioning whether OpenAI accessed his research logs to scoop the prize, which the company denied while claiming significant progress on a second problem. With rumors that Anthropic has solved yet another, three of the six outstanding problems could be solved this year.
The open letter 'A Severe Misalignment of AI in Mathematics' argues the labs' push to solve famous problems as benchmarks is detrimental to the science and community of mathematics — and frames math as a microcosm of 'a general threat to intellectual work' facing every scientific and creative profession. Notably, the signatories acknowledge AI is here to stay and genuinely useful for research; Tao himself has adapted his own process around the tools.
— 'A Severe Misalignment of AI in Mathematics,' open letter led by Terence Tao. The mathematicians' core worry: the 'mass production at faster and faster pace of true/false statements could destroy fertile ground instead of breathing life into new ideas' — answers without the long communal process of talks, discussion, and simplification that turns landmark proofs into human understanding.
Eric Weinstein dismissed the letter as a 'priesthood' defending turf and Taleb called it lost privilege, but Steven Strogatz's response cut deeper: at 67, loving the tools and using them daily, he still got choked up describing the 4,000-year history of mathematics to a reporter. His point — even change you view as inevitable involves loss worth mourning.
A new Census Bureau study covering roughly 29% of bachelor's degrees earned from 2016 to 2024 found graduates entering highly AI-exposed fields saw a five-percentage-point employment drop and a 13% earnings reduction beginning with ChatGPT's November 2022 release — a decline the authors compare to graduating into a large recession, driven partly by grads taking lower-paid roles outside their field.
Companies were slow to adopt AI, so the idea that hiring practices changed the moment ChatGPT launched is suspect; tech layoffs trace back to pandemic-era over-hiring, and other studies find the same effects equally correlated with work-from-home hollowing out junior employees' relationships. Whatever the cause, the graduate job market is brutal — and worthy of continued study.
ZAI announced a $5 billion raise split between new stock issuance and convertible bonds, with 60% earmarked for training next-generation models — including a 'fully self-training loop.' It's the first time a Chinese lab has openly declared it will build RSI.
A 33-researcher paper backed by ByteDance, 'The Last AI Built by Humans: Towards Genuine Recursive Self-Improvement,' maps a theoretical roadmap and concludes the method isn't there yet — AI can automate parts of training but can't yet improve the process itself. Still, the flurry of Chinese RSI ambition reinforces that if recursive self-improvement is the safety community's core concern, pacing the frontier can't be a Western-only project.
