Amid increasing concerns about the rapid development of artificial intelligence, Anthropic CEO Dario Amodei has proposed a set of strategies aimed at slowing the pace of AI progress to enhance safety. In a recent blog post, Amodei emphasized the need to "pace the frontier" of AI capabilities, citing recent events such as the OpenAI-HuggingFace security breach and the accelerated ability of AI systems to improve themselves as key reasons for caution.
Amodei outlined three main approaches. First, he advocates for the introduction of "embedded evaluators"—independent third-party experts who would work inside AI companies to verify adherence to safety commitments and ensure transparency around safety incidents. Anthropic has committed to implementing this measure unilaterally and encourages governments to require similar practices across the industry.
Second, Amodei called for coordination among leading AI companies in democratic countries to establish common safety standards and limits on unchecked AI development. He acknowledged challenges to such collaboration, including concerns about antitrust regulations, and suggested that government facilitation with narrow waivers could help enable these discussions.
Third, he urged global cooperation, including engagement with authoritarian governments such as China, to address the risks of AI misuse. While recognizing the difficulties, Amodei highlighted the importance of agreements to prohibit dangerous applications like AI-enabled biological weapons.
These proposals come in the context of heightened debate over AI safety, including the resignation of Anthropic researcher Jacob Coxon, who expressed alarm over the industry's trajectory. Amodei stressed that his intent is to balance optimism about AI's potential benefits with a commitment to developing the technology responsibly.
The conversation around AI pacing reflects broader concerns about trust in technology companies and regulators. Critics caution that apocalyptic warnings may distract from current harms caused by AI, while others warn of regulatory capture if industry-led safety measures dominate. Amodei concluded by reaffirming his belief in AI's capacity to improve human life, provided development proceeds with deliberate care and oversight.