Anthropic CEO Calls for Slowing AI Development

Anthropic’s chief executive, Dario Amodei, has announced that the company will begin to slow the pace of its AI development and will grant third‑party evaluators, such as METR, broad access to its models. The move is intended to reinforce the company’s commitment to safety practices and to allow independent verification of its compliance with internal and external standards.

Amodei outlined a three‑step plan to “pace the frontier.” The first step, already underway, is the unilateral provision of model access to external reviewers. The second step calls for industry‑wide cooperation, likely involving government partners, to establish common safety benchmarks and oversight mechanisms. While the final step is not detailed in the brief, it is implied to involve further regulatory alignment and coordinated risk mitigation across the sector.

By opening its models to independent scrutiny, Anthropic aims to create a more transparent development cycle and to give regulators the information needed to assess potential risks. The company’s initiative reflects a growing trend among AI firms to prioritize safety and oversight as they scale up model capabilities. For more detail, see the full story on The Verge.