Anthropic CEO proposes three-step plan to slow AI development amid industry divide

Anthropic CEO Dario Amodei is pushing for a structured slowdown of AI development. His proposed three-step plan relies on third-party evaluators within labs, coordination across the domestic industry, and international agreements that may involve government assistance. While some leaders like Sam Altman and Demis Hassabis appeared to agree with aspects of this, the industry remains split.

Adopting such a framework would cost companies their current autonomy over development speed and internal lab access. Meta CEO Mark Zuckerberg has explicitly opposed limiting this autonomy, arguing that individual labs have the incentive and responsibility to move at their own safe pace. This internal conflict is further complicated by reports that Zuckerberg, Elon Musk, and Jensen Huang blocked the creation of a private nonprofit regulator similar to FINRA.

The path to mandatory national safety standards is currently blocked by the executive branch. The Trump administration has labeled the AI safety crisis a hoax and claims that existing criminal and regulatory powers are sufficient to stop bad actors. This creates a environment where some companies seek bipartisan congressional frameworks while the presidency advocates for a lack of specific AI guardrails to ensure the U.S. wins the AI race.