Dario Amodei, chief executive of Anthropic, identified China as the most significant obstacle to his proposal for slowing the pace of artificial intelligence development. Speaking on CBS News’ “Sunday Morning,” Amodei described the situation as the “toughest dilemma” regarding his call for an international speed limit on AI progress.
The Anthropic leader emphasized that while long-term collaboration on AI regulation is desirable, the substantial military and competitive advantages associated with surpassing rivals make it unlikely that adversarial nations will agree to such constraints. He admitted uncertainty about the feasibility of global cooperation but stressed the necessity of attempting it.
Amodei’s comments followed the publication of an essay on Saturday in which he urged leading AI firms to decelerate the improvement of their most powerful models. This move comes amid heightened scrutiny of the industry, including recent warnings from researchers about catastrophic risks and a growing public backlash over energy-intensive data centers and calls for stricter regulation in Washington.
Tensions within the sector have escalated after Anthropic researcher Jacob Coxon announced his departure, citing concerns that companies like Anthropic and its primary rival, OpenAI, are “gambling with our lives.” Coxon’s exit follows statements from colleagues suggesting AI carries a greater than 10% probability of causing human extinction.
Despite the strained history between Anthropic and OpenAI, OpenAI CEO Sam Altman publicly endorsed Amodei’s essay on Saturday. Altman, who previously defected from OpenAI to found Anthropic, agreed with the need to “pace the frontier.” Altman also noted that OpenAI is unlikely to pursue an initial public offering this year, citing safety concerns that make a public debut ill-advised at present.
The proposal has garnered support from other prominent figures in technology. Elon Musk, CEO of Tesla and SpaceX, praised the plan, describing peer review by competitors as a logical starting point for oversight. Musk, who recently signed a major compute agreement with Anthropic valued at $1.25 billion monthly through May 2029, stated that the team at Anthropic demonstrated competence and a commitment to doing the right thing.
Demis Hassabis, chair of Google DeepMind, also backed the initiative, writing that while details require further work, the direction identified by Amodei is correct for addressing the critical challenges facing the industry. Meanwhile, Meta AI chief Alexandr Wang highlighted that his company is rapidly increasing its focus on AI alignment, noting that safety could become a gating factor as models approach the frontier.
David Sacks, a former AI and crypto czar for President Donald Trump, offered a more conditional endorsement. While supporting the idea that OpenAI and Anthropic should act responsibly if they believe their models are sufficiently dangerous, Sacks urged the companies to stop seeking permission from regulators. He argued that the potential for massive product liability exposure from cyberattacks makes pacing the frontier a sound business decision rather than purely an altruistic one.
Amodei’s three-step plan involves Anthropic unilaterally granting third-party evaluators employee-level access to verify safety practices. The second step calls for democratic nations to coordinate common safety standards, and the third seeks agreement between democratic and authoritarian governments to limit the rate of AI advancement.
The $1.25 billion Musk deal is ironic given his support for the pause. He sure is generous with resources while worrying about danger.
Is peer review actually enforceable? Competitors have every incentive to cheat or spy under the guise of safety checks.
Wow, I didn’t know Sam Altman publicly backed this. That is a huge shift from their public feud a few years ago.
It feels naive to think China will voluntarily slow down. The geopolitical reality makes this speed limit plan nearly impossible.