Anthropic CEO Dario Amodei has put forward a three-part proposal aimed at addressing AI safety concerns. The plan follows the resignation of AI safety researcher Jacob Coxon from Anthropic and recent revelations regarding an OpenAI/Hugging Face incident.
Amodei's proposal, detailed in a 3,800-word letter titled “We Must Pace the Frontier,” includes committing Anthropic to having third-party AI system evaluators working within the company with full system access. The second part suggests that frontier AI companies in “democratic countries” establish common safety standards and limits on the rate of unchecked AI progress, with government regulation as needed. The third part aims to include “authoritarian countries” in a broader compact.
Leaders from other major AI companies, including Sam Altman of OpenAI, Elon Musk of xAI, Demis Hassabis of Google Deepmind, and Satya Nadella of Microsoft, have expressed support for Amodei's proposals. Amodei states that “pacing does not mean halting model training or technical progress,” but rather aims to provide one to two years of extra time for research into interpretability, alignment, and better testing methods.