Dario Amodei

Dario Amodei used a social post to argue that framing AI regulation as a choice between concentrated power via regulatory capture or broad distribution is a false dichotomy, saying the impact depends on specific rules and institutional design ¹. He said Anthropic aims to craft policies that slow frontier labs while advantaging smaller competitors, citing support or ambivalence toward California bills with revenue/training-cost carveouts, and advocating more rigorous testing for frontier models through CAISI/White House processes and the Pacing the Frontier approach ¹. He views AI as structurally power-concentrating due to scaling dynamics, with open-weights helping but insufficient, and contends well-designed rules can mitigate risks, constrain frontier labs, and still allow open-weights with tailored safeguards ¹. He endorsed reported federal moves toward pre-deployment testing for frontier and, as they approach the frontier, open-weights models, and expressed support for Demis Hassabis’s FINRA-like idea, contrasting this with the industry stance six months earlier ¹.