Anthropic CEO Dario Amodei proposes oversight for AI development

The proposal follows internal turmoil at Anthropic, where resignations have surfaced over concerns that developers are gambling with existential risks. Amodei cited the rapid acceleration of self-improving AI and recent security breaches as primary motivators for his pivot toward caution. His strategy centers on embedded evaluators from organizations like METR, who would receive physical access and internal data to monitor compliance with safety standards—a model he likens to bank regulation.

Beyond internal oversight, Amodei advocated for government-mediated coordination between leading firms to establish industry-wide limits. While he acknowledged concerns regarding antitrust laws and potential competitive disadvantages, he argued that narrow waivers from the U.S. government could enable these safety discussions without triggering legal retaliation. Regarding global competition, particularly with China, he suggested that restricting access to semiconductor equipment and advanced chips would maintain an American lead while allowing for limited international agreements on specific threats, such as the use of AI in biological weapon production.

The initiative has drawn skepticism from industry critics who view such regulatory proposals as a form of "regulatory capture" designed to insulate established players from competition. Critics argue that focusing on hypothetical, apocalyptic scenarios distracts from the immediate harms current AI tools inflict. Amodei maintains that his stance is a search for a balanced path, asserting that the potential for AI to improve human life remains achievable only if the technology is built with deliberate, verifiable care.