While the proposal has received public support from leaders at OpenAI, Google DeepMind, and SpaceX, Amodei acknowledged that a slowdown is difficult to maintain if adversarial nations, specifically China, do not adopt similar speed limits.
The proposal arrives as AI capabilities rapidly expand, raising concerns among researchers that companies are gambling with safety to maintain a competitive edge.
This tension is heightened by the significant military advantages and commercial incentives associated with leading the field.
While American firms like Anthropic and OpenAI are currently at the forefront, Chinese startups have recently gained momentum, prompting calls from Chinese President Xi Jinping for a global governance framework and the creation of an open-source AI community among BRICS nations.
To address these risks, Amodei outlined a three-step mechanism that begins with Anthropic allowing third-party evaluators to verify its safety practices.
The plan then calls for common safety standards among democratic nations before eventually seeking coordination with authoritarian governments.
This effort to align AI—the process of ensuring systems follow human values—is becoming a critical priority for major firms, though critics suggest the move also serves as a business strategy to limit legal liability and manage the public backlash surrounding AI infrastructure and data centers.