Anthropic CEO Dario Amodei called for slowing AI development. According to TechCrunch AI and The Verge AI, he described a three-stage plan that should give companies time to build safeguards, and regulators time to assess models.
The first step Anthropic named providing external evaluators, including METR, broad access to models to verify compliance with safety practices and obligations. The available description states that this step is undertaken unilaterally by the company.
The materials also note that Sam Altman and Amodei appear to agree on the idea of “slowing down.” However, the full content of the plan, its timelines, and potential commitments from other developers are not clear from the provided data.
editorial commentary
Why it matters
Likely consequence — increased role of external evaluators and more formal procedures before releasing new models. The next observable signal will be the publication of details of the three-stage plan or confirmation of specific commitments by other developers. Substantial uncertainty remains due to limited materials and lack of a primary statement.