Hinton, in a CNN interview, calls for FDA-style regulatory requirements for AI
This story was merged into OpenAI's proposal for binding global AI standards →
Merged·Day 3·in the edition since 24 Sep 2026·5 sources·updated 26 Sep 2026
Ball's call for AI development pacing
The gist
Amodei Seeks Antitrust Waiver for AI Pacing; Schmidt, Bessent Dissent
The debate involves the CEOs of the largest frontier AI labs, a former major tech executive, and a sitting cabinet secretary, centering on whether coordinated AI development limits are prudent policy or anticompetitive protection. The antitrust waiver request is a concrete legal mechanism being proposed to enable industry coordination that is currently legally constrained.
The full picture
Anthropic CEO Dario Amodei published 'We Must Pace the Frontier' on September 12, 2026, requesting an antitrust waiver to allow competing frontier AI labs to jointly discuss limits on their development pace. Sam Altman of OpenAI, Elon Musk of xAI, and Demis Hassabis of Google DeepMind each endorsed variations of the idea within hours. Dean Ball issued a more specific call, proposing that pacing begin on November 19, 2026, for a period no shorter than 30 days. The slowdown envisioned by pacing letter signatories is described by Mowshowitz as temporary and still faster than current AI progress rates.
Former Google CEO Eric Schmidt rejected the idea, telling The Economist's Zanny Minton Beddoes that 'we're not going to do a pause, because the entire incentive system in America is not to pause,' and framing AI development as a race in which the choice is whether people want their lives mediated by an American or Chinese AI model. Treasury Secretary Scott Bessent separately argued that the US needs stronger open-source AI models, warning that large AI labs could achieve regulatory capture that would stifle innovation, and framing open-source models as a strategic counter to China.
The regulatory capture question sits at the center of the debate. Critics argue that labs advocating pacing seek competitive protection. Mowshowitz disputes this, arguing genuine pacing would slow leading labs more than smaller entrants. Roon, identified with OpenAI, argues that leading labs bear significant costs to understand AI capabilities and safety first, creating a blueprint later entrants can follow more cheaply, and predicted open-source AI will eventually be banned after a major disaster. Yo Shavit argued pacing efforts should focus narrowly on alignment at the frontier rather than misuse restriction, because the latter would be politically intractable.
How it developed
Obama argues agentic AI deployment is driven by commercial imperatives misaligned with societal needs.
US rejection of global AI governance
White House science adviser Kratsios rejected any global governance scheme for superintelligence at the UN Security Council on September 23, urging nations to build their own technical capacity instead.
Trump's September 22 UN address declared the US leads China in AI 'by a lot' and designated the Department of Justice as the oversight mechanism. Epoch AI's JS Denain estimated Chinese labs trail US frontier labs by six to eight months on public releases based on Epoch's ECI metric, with some benchmarks placing the gap closer to two to four months.
First in the edition · 24 September 2026
Law360 reported that Altman, Musk, and Hassabis each endorsed variations of Amodei's antitrust waiver proposal within hours of its publication
Dean Ball calls for pacing beginning November 19, 2026 for no fewer than 30 days
Srinivas argues on 20VC that US export controls, not technical disadvantage, explain China's capability gap, and may be producing a stronger long-term competitor
PIIE publishes analysis assessing Trump-Xi summit as unlikely to produce meaningful AI governance progress.
Roon predicts open-source AI will be banned after a major disaster; regulatory capture debate intensifies
Ball's call for AI development pacing
Mowshowitz clarified September 26 that Ball's proposed 30-day pacing of frontier AI, to begin November 19, targets a rate still faster than today's progress and is temporary; he identified a structural barrier: people unconvinced AGI is near resist bearing any caution costs.
Reporting on Trump's statements argued media mischaracterized him as flatly opposing a slowdown; Trump accepted 'guardrails' while opposing a moratorium, his administration had already slowed two model releases, and the probability of a federal AI safety bill fell to 18%.
Schmidt argues financial scale, not energy, is the binding limit on AI infrastructure expansion.
Mowshowitz publishes 'The Pacing of the Frontier,' clarifying signatories envision a temporary slowdown still faster than current rates
Sources
Rohan Paul Twitter
Dean Ball Twitter
Zvi's AI Roundups
thezvi.substack.com
Related
The daily email
Want this in your inbox?
I send one email each morning with the stories that moved. If you would rather just read here, that works too.
Subscribe free