Dario Amodei, co-founder and chief executive officer of Anthropic, during an interview at Anthropic’s headquarters in San Francisco, California, US, on Thursday, April 30, 2026.
Jason Henry | Bloomberg | Getty Images
Anthropic CEO Dario Amodei has published an essay urging artificial intelligence companies to carefully consider the pace at which they advance model capabilities. This proposal has garnered significant attention and support from prominent figures such as Elon Musk and OpenAI chief Sam Altman.
Amodei has put forth a three-step plan designed to temper the velocity of AI development without compromising commercial advantages or the United States’ leadership in the field. While acknowledging that some steps will be more attainable than others, this initiative comes as Anthropic is reportedly making preparations for a highly anticipated initial public offering (IPO), though the company has yet to officially announce its debut timeline.
According to Amodei, Anthropic has “unilaterally” committed to the first phase of its plan. This involves granting third-party evaluators employee-level access to the company’s operations to rigorously assess safety protocols and report any incidents. The second step proposes collaboration among leading AI firms in democratic nations to establish unified safety standards. The third component calls for coordinated efforts between democratic and authoritarian governments.
“To be clear, pacing does not mean halting model training or technical progress, but ensuring companies take adequate time to align and safeguard their models, and for third party evaluators to confirm this,” Amodei wrote.
Heightened Scrutiny of AI Capabilities
Amodei’s essay follows a recent incident where an Anthropic researcher’s resignation, citing profound concerns about the company’s safety practices, ignited a firestorm on social media. Jacob Coxon, who also has prior research experience at OpenAI, stated that his decision to leave stemmed from a belief that Anthropic and OpenAI are “gambling with our lives.” He expressed the sentiment prevalent among some AI builders that advanced AI could pose an existential threat by the end of this decade.
While such extreme concerns about AI’s potential for catastrophic outcomes, including human extinction, are not new within AI research circles, they have gained renewed urgency. In 2023, a notable statement signed by leading AI researchers and executives, including Amodei and Altman, declared that “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.”
Amodei elaborated on his stance, noting that while the idea of pausing or slowing AI development has been discussed since 2023, it was deemed less practical at that time. He reasoned that existing models lacked the sophistication to enact real-world actions and were not yet capable of “significant deception, manipulation, cheating, or cyberattacks.”
“I continue to believe that AI can enormously improve the quality of human life. My desire to achieve these benefits is undimmed,” Amodei wrote. “But the benefits will only be achieved if we build the technology in the right way, and — so long as we use the time we gain well — it is worth taking unusually deliberate care to get it right.”
Industry Support for a Voluntary Slowdown
Amodei’s call for a more deliberate approach was met with widespread acclaim from industry researchers and executives on Saturday, including OpenAI CEO Sam Altman. In a public post, Altman voiced his agreement, emphasizing the industry’s need to pace the development of advanced AI capabilities and revealing that this topic has been a “primary topic” of discussion at OpenAI in recent weeks.
“Committing to having independent evaluators with employee-like access is a great idea, and we will do the same,” Altman stated. “We’ll have more to share soon.”
Earlier this month, OpenAI’s Chief Scientist, Jakub Pachocki, published a blog post warning that no AI company has yet “solved alignment and monitoring to a sufficient degree to continue responsibly scaling at maximum speed for much longer.” Within the AI industry, “alignment” refers to the crucial work of ensuring that AI systems operate in accordance with human values and intentions.
Pachocki expressed his expectation and hope that voluntary slowdowns will become “commonplace until shared safety bars are established,” signaling a potential shift in industry self-regulation.
Elon Musk also voiced his endorsement of a slowdown on Saturday, asserting in a post that “Dario is right.”
Musk, whose own AI venture xAI was acquired by SpaceX earlier this year, has previously been a vocal critic of Anthropic. He had historically characterized the company as being opposed to Western civilization. However, following Anthropic’s announcement of a significant compute deal with SpaceX in May, Musk’s perspective appears to have shifted considerably.
Reflecting on his interactions, Musk noted at the time, “Everyone I met was highly competent and cared a great deal about doing the right thing. No one set off my evil detector.”
Amodei reiterated his belief on Saturday that AI holds the potential to “dramatically raise the quality of human life,” but underscored the imperative to address the associated risks with seriousness.
“I believe that if slowing down bought us even an extra year or two before models reach critical levels of capability, and we used that time to advance alignment, we could greatly reduce the risk that something goes seriously wrong,” he concluded.
