We Must Pace the Frontier · 2026-09-12 · primary
Themes: US-China Framing · Closed vs. Open Split · Safety Debate
Compute, not algorithms or talent, decides who can compete at the frontier.
- Do not sell powerful AI chips or semiconductor manufacturing equipment to China, and crack down on chip smuggling operations and remote access to data centers outside China. Chips will be the main determinant of China’s AI strength. - Crack down on unauthorized [distillation](https://www.cisa.gov/news-events/cybersecurity-advisories/aa26-251a) by companies in authoritarian countries. Distillation of frontier models allows lagging companies to narrow the gap using a fraction of the cost it would take to develop their own AI independently.
We Must Pace the Frontier · 2026-09-12
Restricting chips and stopping industrial-scale distillation is the effective way to slow a rival's frontier, and licensing models is not.
Companies and the US government should cooperate to make these steps as effective as possible. Anthropic has [consistently](https://darioamodei.com/post/on-deepseek-and-export-controls) [advocated](http://wsj.com/opinion/trump-can-keep-americas-ai-advantage-china-chips-data-eccdce91) for all of these measures, because we’ve always understood that they would be essential to any pacing.
We Must Pace the Frontier · 2026-09-12
If we execute these measures well, I believe they would slow China’s progress enough to widen America’s lead significantly over the next 3–5 years — the window when AI becomes geopolitically most important.
We Must Pace the Frontier · 2026-09-12
Capable models should be required to pass pre-release testing, binding open weights and closed APIs alike.
Broadly speaking, I am most enthusiastic about pacing based on what a given frontier AI system can _do_, and how safe we observe it to be. For example, one possible scheme might be a series of “checkpoints”: if models have capability X, then they need to be accompanied by certifications of alignment properties Y and Z — such as some combination of evaluations, interpretability analyses, and audits of training environments — which demonstrate their alignment properties.
We Must Pace the Frontier · 2026-09-12
Testing and evaluation of AI models becomes more difficult as they increase in capabilities. More intelligent models are more capable of deceiving tests, and thus may _appear_ aligned while having serious problems that go undetected. Building up a much broader and more ingenious stable of evaluations, along with interpretability analysis to cross-check them, would be hugely valuable, and a lot of progress could be made on this in 1-2 years.
We Must Pace the Frontier · 2026-09-12
An agreement by both sides to test their models before release for acute risks in areas such as cybersecurity, biology, and alignment. As noted above, this could be done through a global standards body. I actually think creating such a body is likely feasible, but giving it real teeth will be a challenge, and the difficulty will be in verification that both sides don’t have secret models which they don’t test but may deploy in secret (e.g., for military applications). -
We Must Pace the Frontier · 2026-09-12
A coordinated slowdown is worth pursuing even at the cost of advantage, because uncoordinated speed is the greater risk.
The idea of pausing or slowing AI has been floated [as far back as 2023](https://futureoflife.org/open-letter/pause-giant-ai-experiments/), and I think it made little sense back then. The question was always: _what would you do with the extra time_? The AI models of those days were not powerful enough to act as agents in the world in any coherent way, and were not capable of significant deception, manipulation, cheating, or cyberattacks. Slowing down in order to address their alignment risks felt like trying to study the psychology of humans by performing experiments on bacteria.
We Must Pace the Frontier · 2026-09-12
But over the last few months, I have become convinced that fully addressing the risks requires even more prudence — not just investing in risk prevention, but pacing the rate of capabilities advancement so that risk prevention has time to keep up. **We must slow the pace at which we improve the capabilities of AI models.** **Progress will still seem fast, and we must make wise use of the time we gain.** Two things have convinced me.
We Must Pace the Frontier · 2026-09-12
Pacing within democracies will be limited by the lead that US companies have over authoritarian regimes, chiefly the Chinese Communist Party. If we slow down by more than this amount, then (unpaced) CCP-associated projects will pull ahead, creating significant national security risk. I agree with [Secretary Bessent](https://www.bloomberg.com/news/articles/2026-09-09/bessent-warns-nothing-would-matter-if-china-wins-the-ai-race) that a Chinese lead in AI would pose grave danger for the United States and the world.
We Must Pace the Frontier · 2026-09-12
Safety and innovation trade off against one another: what makes a model safer makes it less capable or less widely available, so pushing on one goal costs ground on the other.
Along with my co-founders and employees, I have grappled with this duality of risk and benefit since the beginning of Anthropic. Not building the technology deprives humanity of benefits or simply places AI in the hands of authoritarian powers, while building it too fast is reckless. We have sought a middle way: to show that it’s possible to build carefully and succeed commercially, and to make safety something on which AI companies compete. In other words, to create a _race to the top_.
We Must Pace the Frontier · 2026-09-12
But like many technologies before it, AI brings risks, and because it is such a powerful technology, these risks are serious. I’ve [written](https://darioamodei.com/essay/the-adolescence-of-technology) a lot about them too. They include the risk of [losing control of AI systems](https://www.anthropic.com/news/improving-alignment-security-efforts), [misuse of AI for cyberattacks and bioterrorism](https://www.anthropic.com/threat-intelligence-report-september-2026), and [serious economic disruption](https://www.anthropic.com/institute/econ-scenarios).
We Must Pace the Frontier · 2026-09-12
Positions are read from the claim graph, and every passage is fetched from the archive when this page is built. A passage the archive cannot supply fails the build.