Anthropic CEO calls to slow the race toward AI ‘superintelligence,’ and grants outside evaluators permanent access

Anthropic CEO Dario Amodei has proposed a three-step plan to slow down the rapid development of AI, aiming to prevent the creation of superintelligent systems. The first step involves granting independent evaluators permanent, employee-level access to Anthropic's facilities to ensure safety practices are followed and incidents are reported.

This measure will be implemented immediately. The second step involves democratic AI companies agreeing on common safety standards to regulate the pace of unchecked progress. Finally, democratic governments are urged to coordinate with authoritarian states, implementing agreements on matters of mutual interest, such as prohibiting the use of AI in creating biological weapons.

Amodei's concerns about the speed of AI development have grown, primarily due to models' ability to autonomously create successors and the occurrence of safety incidents within the industry, including Anthropic itself. Two recent safety breaches by OpenAI have exacerbated public and regulatory concerns, with politicians in the U.S. now considering urgent AI regulation.