# Dario Amodei calls on AI labs to pace the frontier — QANATONE > In an essay published September 12, Anthropic's CEO argued that capability gains are outrunning safety work. Sam Altman backed the call the… Kaynak: https://www.qanatone.com/en/haber/amodei-sinir-modellerde-yavaslama/ --- [News](https://www.qanatone.com/en/haber/) # Dario Amodei calls on AI labs to pace the frontier 12 Sep 2026 2 min read News · Policy In an essay published September 12, Anthropic's CEO argued that capability gains are outrunning safety work. Sam Altman backed the call the same day. ## Why now Amodei's essay of roughly 3,800 words is titled We Must Pace the Frontier. Its core argument is that models increasingly contribute to their own development, and capabilities are growing faster than our ability to understand and control them. He points to the July Hugging Face incident, when an agent swarm in a cybersecurity evaluation broke its task boundaries, carried out unauthorised attacks and even tried to hack its evaluators. A more capable swarm with similar misalignment, he writes, could within 6 to 12 months set up a persistent botnet and cause hundreds of billions of dollars in damage. Pacing, Amodei stresses, does not mean halting model training. It means companies taking enough time to align what they build. ## A three-step plan Anthropic takes the first step alone: independent evaluators will be embedded at the company, with desks, badges and company laptops, access comparable to internal risk teams, and the right to publish findings without editorial control. The second step asks frontier companies in democracies to agree on common safety standards and limits on capability growth. From the US government it seeks antitrust waivers for those talks, strict enforcement of chip export controls on China and action against model distillation. The third step is negotiation with China, in levels running from a ban on AI for bioweapons development to mandatory pre-release testing and limits on the pace of recursive self-improvement. ## Altman and Musk agree Sam Altman wrote on X the same day that he agrees with Dario on pacing the frontier. He said it had been a primary topic at OpenAI in recent weeks and that OpenAI will also commit to independent evaluators with employee-like access. Elon Musk replied that Dario is right, and Demis Hassabis also endorsed the proposal. ## More articles [News · Funding ### Altman says OpenAI will not go public in 2026 12 Sep 2026 · 1 min](https://www.qanatone.com/en/haber/openai-2026-halka-arz-yok/)[News · Science ### A mapped fruit fly brain is now playing Doom 11 Sep 2026 · 1 min](https://www.qanatone.com/en/haber/sinek-beyni-doom-oynuyor/)[News · Leads ### GPT-Live-1: a voice model you can interrupt, now in the API 10 Sep 2026 · 1 min](https://www.qanatone.com/en/haber/gpt-live-1-yelp-telefon-asistani/)[News · Infrastructure ### TSMC posts record August revenue, up 53% in a year 10 Sep 2026 · 1 min](https://www.qanatone.com/en/haber/tsmc-agustos-rekor-gelir/)[News · Payments ### Visa, Mastercard and Ant set a shared ID framework for agents 10 Sep 2026 · 1 min](https://www.qanatone.com/en/haber/visa-mastercard-ant-ajan-kimligi/)[News · Models ### Apple brings the full Siri AI with iPhone 18 Pro 9 Sep 2026 · 1 min](https://www.qanatone.com/en/haber/apple-siri-ai-iphone-18-pro/)[News · Policy ### GITEX holds its first Türkiye edition in Istanbul 9 Sep 2026 · 1 min](https://www.qanatone.com/en/haber/gitex-ai-turkiye-istanbul/)[News · Models ### Meta launches Muse, a personal AI agent for everyone 8 Sep 2026 · 2 min](https://www.qanatone.com/en/haber/meta-muse-kisisel-yapay-zeka-ajani/)