Anthropic CEO Dario Amodei Outlines Plan to Slow AI Development

Anthropic CEO Dario Amodei stated that the rapid progress in AI models needs to be slowed down, announcing three strategic steps that include third-party evaluators and coordination among democratic countries.

A New Approach to Artificial Intelligence Development

In a published blog post, Anthropic CEO Dario Amodei emphasized that the pace of developing AI model capabilities needs to be slowed down. Amodei stated that adopting a more cautious approach is essential due to the rapid advancement of the technology and safety concerns.

Independent Third-Party Evaluators

The first step proposed by Amodei involves embedding evaluators from third-party organizations such as METR inside the company. These evaluators will audit whether companies comply with their safety commitments, and Anthropic announced that it will unilaterally commit to this practice.

Coordination Among Democratic Countries

As a second strategy, it was stated that leading AI companies should coordinate on common safety standards and limits on the speed of progress. To overcome antitrust concerns, it was proposed that the US government should act as an intermediary by granting exemptions for these discussions.

Global Cooperation and Collaboration with China

In the final stage, it was stated that the US and its allies should establish global coordination with authoritarian governments like China in narrow and specific areas, such as banning dangerous uses like the production of biological weapons.