Amodei's pacing plan caps the slowdown at the size of America's lead
Neither Washington nor Beijing endorsed the proposal behind this month's frontier-lab consensus, which asks for evaluators inside labs, coordinated capability thresholds and international deals and ties how much anyone slows to how far ahead the US already is.
The Engineer · Build desk
What happened
- Dario Amodei, Sam Altman and Elon Musk publicly endorsed a plan to slow frontier AI development, an agreement among rivals that lasted about a week before both relevant governments answered it.
- Trump phoned Nvidia CEO Jensen Huang live on stage at a tech summit to hear him agree that a slowdown "is not going to happen." On Truth Social he called warnings about AI destroying humanity a "HOAX."
- Asked about Amodei's proposal on September 14, Chinese Foreign Ministry spokesperson Guo Jiakun criticised "fearmongering, confrontation and vicious competition," according to the Associated Press. He did not commit China to slowing its own development.
- Cybersecurity stocks rose double digits this week as investors bet that AI-capable hacking is the nearer risk than the superintelligence scenarios under debate.
Why it matters
- What it limits constraint The coordination layer cannot start without government help clearing antitrust barriers. Shared capability thresholds among US labs are gated on a permission only Washington can issue.
- Who has to decide decision Pacing as defined leaves training and technical progress running, so teams sequencing work around an imminent capability freeze have misread the essay.
- What doesn't add up contradiction Beijing's rejection is aimed at who writes the rules: its own standards and cyberspace regulators say AI development must be monitored because it may escape human control.
- Who's exposed exposure The restraint that changes an engineering team's tool choice this quarter is procurement policy, with at least three major companies reported to be limiting staff use of one leading assistant on sensitive work.
Next story · Build · Sep 29
White House super intelligence accord leaves enforcement to the six firms that signed itConfirmed5 publishers
Pacing, as Amodei writes it, still allows training and technical progress [6]. The essay asks for three things: permanent independent evaluators embedded inside frontier labs, and coordination among leading companies in democratic countries on shared safety standards and capability thresholds. The third is international agreements covering at least some dangerous capabilities [6][8]. How much pacing is permitted depends on one variable. Democracies can slow only by roughly as much as their lead over geopolitical rivals allows [15]. A larger lead, Amodei wrote, gives democracies the "breathing room we need in order to pace effectively" [16]. So the plan's first ask is a bigger lead: tougher restrictions on advanced chips and chipmaking equipment, on unauthorized model distillation, and on theft of model weights. Amodei believes those could widen the American advantage over the next three to five years [7].
The second layer is the one with a dependency outside the labs. Coordination on shared thresholds may need government help resolving antitrust barriers [8]. Asked whether AI companies should slow development or face more regulation, Trump pointed at China. "We're leading China in AI," he said. "Whoever wins AI wins" [2].
Beijing answered in two registers that do not carry the same weight. The Neuron notes that a Foreign Ministry statement is an official government position, while a Global Times commentary signals political thinking and binds nobody. That commentary called the proposal a "Cold War playbook" and argued its restrictions on China would entrench US dominance [14][5]. China Daily's objection was procedural. The three-company arrangement, it wrote, "sounds rather like a club whose membership rules have been drafted before the guest list is announced. A global AI-safety framework that excludes China is not quite global" [17]. Chinese regulators have said the underlying risk is real: the Standardization Administration of China, working with the Cyberspace Administration of China, says AI development must be monitored as it may go beyond human control [11].
The warning that started this came from Jacob Coxon, who quit Anthropic. Frontier labs, he said, are moving toward increasingly autonomous systems before anyone knows how to reliably control much more capable versions of them [9]. His concern centers partly on recursive self-improvement, and OpenAI says its own agents are already taking over larger pieces of the research process [9][18]. A roadmap from more than thirty Chinese researchers sorted self-improvement into five levels and surveyed nearly 500 papers, finding under 6% at the top tier. It flagged coding as the domain most likely to reach that tier first, since fixes there can be tested instantly [19]. Six percent of 500 is 30, so the top-level literature runs to a few dozen papers at most [26].
Products kept shipping through the argument. Apple began rolling out its rebuilt Siri, built in part on Google's models and absent from the EU and China at launch [22]. Salesforce shipped an in-house reasoning model trained entirely on synthetic data to keep customer information away from outside AI providers [23]. According to China Daily, the pressure on frontier spending is commercial: users are shifting demand toward cheaper tokens such as Kimi K3 (low) and DeepSeek V4 Pro over expensive frontier models [24].
The one text this week that would actually constrain a model's behavior is Microsoft's draft "Humanist AI" code of conduct, which would require its models to accept being shut off and forbid them from editing their own reasoning logs or reasoning in unreadable shorthand. It is open for public comment, and for now it binds only Microsoft's own systems [12].
What to watch
- Whether the administration grants any antitrust relief that would let rival US labs compare capability thresholds.
- Whether the US-China AI-safety dialogue China Daily pointed to produces anything binding on either side.
- Whether Amodei's three-to-five-year lead-widening forecast holds through the next round of cheap Chinese model releases.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence78
- Adoption22
- Hype gap+34
- Incentives82
- Confidence74
Perspective Coverage
7 publishers- Builder
- Builder 33%
- Operator
- Operator 35%
- Investor
- Investor 32%
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
President Trump used Truth Social to call warnings about AI "taking over the World, destroying Humanity" a "HOAX," comparing it directly to climate change.
- [2]
Asked whether AI companies should slow development or face more regulation, Trump said: "We're leading China in AI. Whoever wins AI wins."
- [3]
Dario Amodei, Sam Altman and Elon Musk publicly endorsed a plan to slow frontier AI development.
- [4]
Trump phoned Nvidia CEO Jensen Huang live on stage at a tech summit just to hear him agree that a slowdown "is not going to happen."
- [5]
The state-run Global Times commentary on Amodei's essay called the proposal a "Cold War playbook," arguing that its restrictions on China would entrench U.S. dominance.
- [6]
In "We Must Pace the Frontier," Amodei's plan includes independent evaluators inside frontier labs, coordination among American companies, and efforts toward international agreements; pacing, he writes, would still allow training and technical progress.
- [7]
Amodei calls for restricting China's access to advanced chips and chipmaking equipment, combating unauthorized model distillation, and preventing model-weight theft, and believes those measures could widen America's lead over the next three to five years.
- [8]
The second layer of Amodei's proposal wants leading AI companies in democratic countries to coordinate around shared safety standards and capability thresholds, potentially with government help resolving antitrust barriers.
- [9]
Former OpenAI and Anthropic researcher Jacob Coxon quit Anthropic and warned that frontier labs are moving toward increasingly autonomous AI systems before anyone knows how to reliably control much more capable versions of them; his concern centers partly on recursive self-improvement.
- [10]
On September 14, Chinese Foreign Ministry spokesperson Guo Jiakun criticized "fearmongering, confrontation and vicious competition" in response to a question about Amodei's proposal, according to the Associated Press; his remarks offered no commitment to restrain Chinese AI development.
- [11]
The Standardization Administration of China, in cooperation with the Cyberspace Administration of China, says that the development of AI technology must be monitored as it may go beyond human control.
- [12]
Microsoft's AI division published a draft "Humanist AI" code of conduct that would require its models to accept being shut off, forbid them from editing their own reasoning logs or reasoning in unreadable shorthand, and explicitly deny them anything resembling rights or personhood; it is open for public comment and for now only binds Microsoft's own systems.
- [13]
China Daily said the planned AI-safety dialogue between China and the US and future high-level exchanges offer opportunities for practical engagement.
- [14]
A Foreign Ministry response is an official government position; a state-media commentary signals political thinking without establishing a binding policy.
- [15]
Amodei says democracies can only slow down by roughly as much as their lead over geopolitical rivals allows.
- [16]
In Amodei's words, a larger lead gives democracies the "breathing room we need in order to pace effectively."
- [17]
China Daily wrote: "The proposed coordination among the three companies sounds rather like a club whose membership rules have been drafted before the guest list is announced. A global AI-safety framework that excludes China is not quite global."
- [18]
OpenAI says its own agents are already taking over larger pieces of the AI research process.
- [19]
More than thirty Chinese researchers published a roadmap sorting AI systems into five levels of self-improvement; surveying nearly 500 existing research papers, they found under 6% reach the top tier, and flagged coding as the domain most likely to get there first since its fixes can be tested instantly.
- [20]
Wall Street sent cybersecurity stocks up double digits this week on the bet that AI-capable hacking is the more immediate risk.
- [21]
At least three major companies are reportedly restricting how their staff use one leading AI assistant on sensitive work over data-retention concerns.
- [22]
Apple began rolling out its rebuilt Siri, which can read what is on screen and act across apps including WhatsApp and Audible, built in part on Google's models and absent from the EU and China at launch.
- [23]
Salesforce rolled out its own in-house reasoning model trained entirely on synthetic data to keep customer information away from outside AI providers.
- [24]
According to China Daily, cheaper models have shifted AI user demand toward cheaper tokens such as Kimi K3 (low) and DeepSeek V4 Pro over expensive frontier models including Fable 5.1, GPT 5.6 Sol, Grok 4.6 and Kimi K3 (max).
- [25]
China's "AI Plus" initiative, announced in August 2025, sets goals for integrating AI across six major areas by 2027 and targets adoption of next-generation intelligent terminals and AI agents above 70% by then, rising above 90% by 2030; these are deployment goals, not limits on training the most powerful models.
- [26]
Under 6% of nearly 500 surveyed papers is fewer than about 30 papers at the top self-improvement tier.
Sources
7 independent publishers whose own reporting we read for this story.
- dev.toThe AI Slowdown Nobody's Actually Slowing Down For
1 article · September 16, 2026
- economictimes.indiatimes.comTrump AI Development: Trump downplays the need to check AI development and says he doesn't want to cede edge to China - The Economic Times
2 articles · September 14, 2026
- mezha.netПекін назвав заклик сповільнити розвиток ШІ тактикою холодної війни
3 articles · September 14, 2026
- techrepublic.comTrump on AI Extinction: Beating China Is the Concern
1 article · September 11, 2026
- the-decoder.comChina fires back at U.S. AI safety warnings, calling them fearmongering to lock in American advantage
2 articles · September 14, 2026
- theneuron.aiAI Safety’s New Paradox: America May Need a Bigger Lead to Slow Down
2 articles · September 14, 2026
- tomshardware.comChinese state media counters Anthropic's call to put brakes on AI development — paper says move is ‘a response to Chinese competition’
1 article · September 16, 2026