Concerns about the rapid advancement of artificial intelligence have prompted Anthropic CEO Dario Amodei to outline a plan aimed at slowing AI development to ensure safety. In a recent blog post, Amodei emphasized the need to "pace the frontier" of AI progress, citing recent security incidents and the accelerating capabilities of AI systems as reasons for caution. He proposed three main strategies: embedding third-party evaluators within AI companies to monitor compliance with safety commitments, coordinating common safety standards and limits on unchecked AI progress among leading AI firms in democratic countries, and pursuing global cooperation, including with authoritarian governments, to restrict dangerous uses of AI.
Amodei announced that Anthropic will unilaterally commit to hosting embedded evaluators from independent organizations, granting them access similar to internal risk teams. OpenAI CEO Sam Altman responded positively, indicating OpenAI will adopt similar measures. The proposal also calls for government involvement to facilitate safety discussions without triggering antitrust concerns, and suggests that export controls on advanced chips and manufacturing equipment could help maintain a technological lead over China.
The call for a more cautious approach follows heightened debate within the AI community, including the resignation of an Anthropic researcher over safety worries. While some critics question the feasibility and motivations behind these proposals, Amodei stressed the importance of building AI responsibly to realize its potential benefits. He described the current skepticism toward AI development as a "crisis of trust" and urged deliberate care in advancing the technology.
The discussion reflects growing industry awareness of AI risks and the challenges of balancing innovation with safety and ethical considerations.