API
< Back to all clusters
[TECHNOLOGY] · United States · 20 sources

started · updated

Anthropic AI development accelerates despite CEO calls for slowdown

An independent assessment by the nonprofit Model Evaluation and Threat Research (METR) reveals that Anthropic is improving its AI model capabilities for well-specified tasks at approximately twice the rate observed between March 2024 and October 2025. The study, which involved an 11-week review of internal data and staff interviews, suggests that this acceleration is largely driven by Anthropic using its own AI systems to assist in research and development, including programming and experiment analysis.

The findings come as Anthropic prepares for a potential initial public offering later this month, which could be one of the largest in history. This development occurs despite a September 12 essay by Anthropic CEO Dario Amodei, in which he urged the AI industry to slow the pace of model improvement to allow safety measures to catch up.

METR noted that while AI is speeding up research, it has not yet reached a stage of fully autonomous contribution to the most critical training phases. Additionally, the report highlighted that Anthropic has recently completed training three internal models that are not yet available to the public. Anthropic has stated that it will soon publish its own analysis regarding its AI progress and Responsible Scaling Policy status.

Entities

Amazon · Anthropic · Claude · Dario Amodei · Google · METR · Microsoft · Sam McCandlish · San Francisco

Claims

What the coverage asserts, and how many sources carry each claim.

Sources