Anthropic CEO Dario Amodei published an essay on Saturday calling for the artificial intelligence industry to intentionally moderate the speed at which it advances model capabilities. The proposal emerges amid rising alarms from researchers who argue that the current pace of development poses existential risks.
Amodei outlined a three-stage approach designed to temper development timelines while preserving both commercial viability and U.S. leadership in AI. He acknowledged that some components of the plan may face greater implementation hurdles than others. The suggestion coincides with growing speculation that Anthropic, which has not yet set an official date for its anticipated initial public offering, is preparing for a major public debut.
According to Amodei, Anthropic has already taken unilateral action to satisfy the first phase of the framework. This involves granting independent third-party evaluators employee-level access to the company, allowing them to audit safety protocols and report any incidents. The second phase encourages leading AI firms in democratic nations to collaborate on establishing shared safety benchmarks. The third phase calls for diplomatic coordination between governments of democracies and authoritarian states.
“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.
The essay was released shortly after significant turmoil within Anthropic. Jacob Coxon, a researcher who previously worked at rival OpenAI, announced his resignation this week, sparking intense debate on social media. Coxon stated he left because he believes Anthropic and OpenAI are “gambling with our lives,” citing internal fears among AI builders that catastrophic outcomes could occur by the end of the decade.
While such existential concerns may appear extreme to the general public, they are well-established within AI research communities. In 2023, prominent figures including Amodei and OpenAI CEO Sam Altman signed a statement asserting that mitigating extinction risks from AI should be treated as a global priority on par with pandemics and nuclear war.
Amodei noted that while pauses on AI development were discussed as early as 2023, he believed that at the time, the technology lacked the power to enact real-world actions or perform sophisticated deception, manipulation, and cyberattacks. He reaffirmed his commitment to the technology’s potential, writing, “I continue to believe that AI can enormously improve the quality of human life. My desire to achieve these benefits is undimmed.”
Leave a Reply