AI: Bosses to Regulate, Politics Resists

A three-step framework proposed by Dario Amodei reignites the debate over the pace of AI. OpenAI, DeepMind, and Elon Musk support the idea of regulating the "frontier" without halting research. On the political side, Donald Trump, JD Vance, and Mike Johnson oppose any pause, citing competition with China, while Bernie Sanders backs a call for a pause.

Trump, Vance, and Johnson Reject a Moratorium, Citing China

Donald Trump dismissed the idea of slowing down AI, asserting that the only safeguard needed is a strong and intelligent president, which the United States has. He claims that his administration prevented AI actors from doing bad or potentially harmful things and argues that authorities already have significant criminal and regulatory power over these companies. He denounces a "maladaptive conspiracy" against AI and data centers, insists that the United States is ahead of China and others and will remain so, and issues a warning to "conspiracy theorists, traitors, and leaks."
JD Vance finds it "a bit strange" that leading AI companies are seeking government regulation, seeing it as a potential "Trojan horse." He recalls the presidential assertion that China is engaged in a long-term race against the United States in AI.
Mike Johnson advocates for a meeting between policymakers and platform actors to find the right balance. He rules out a moratorium, arguing that China would then surpass the United States, and positions the issue between national security and the immediate safety of models. No specific date has been given, although he stated he would hold it "tomorrow."

Anthropic Details Three Steps and Commits to Third-Party Evaluators

Dario Amodei presents a three-pronged framework for regulating AI development. He first proposes integrated third-party evaluators to verify compliance with safety practices and report incidents. He then calls for coordination among leading AI companies in democratic countries on standards and limits. Finally, he advocates for global coordination between democratic and authoritarian governments on the pace of development, "wherever possible." He indicates that Anthropic is unilaterally committed to implementing the first of these steps.

OpenAI Ready for a Federal Framework and Independent Audits

Sam Altman supports the idea of regulating the frontier of AI and reports that the topic has been a major discussion at OpenAI for weeks. He announces the establishment of independent evaluators within the company. On Sunday, he added that OpenAI welcomes a federal framework that would set consistent safety requirements for advanced AI. He hopes other companies will draw inspiration from existing approaches and formulate their own, believing that shared standards for alignment, oversight, and safety will yield better outcomes. He clarifies that regulating does not mean stopping and that progress should proceed more slowly than it could. He identifies two major risks: losing control of the future to AI and excessive concentration of power, which he believes can be avoided by following a "narrow path."

DeepMind Approves the Proposed Direction and Refers to Its Standards Framework

Demis Hassabis believes that Dario Amodei's essay indicates the right direction for this critical moment, while noting that the details remain to be defined. He refers to a framework published in July for an AI standardization body. Elon Musk shares Amodei's view, stating that "Dario is right."

Slowing Down Without Stopping: The Proposed Definition and Scope of the Debate

Dario Amodei clarifies that regulation is not aimed at stopping training or technical progress, but rather at allowing time to align and secure models, under the oversight of third-party evaluators. His essay, published Saturday morning, calls for a structured slowdown of AI development. Following this publication, tech leaders and political figures quickly took positions for or against these ideas. Bernie Sanders sided with a call to "pause AI for the good of humanity," articulated by columnist Peggy Noonan.