Watchdog METR finds Anthropic's AI progress has doubled in speed
Written by AI from abcnews.com's report. Not checked by a person.
The nonprofit METR reviewed Anthropic's internal data and estimated the company is improving its models about twice as fast as a year ago. It attributes the speed-up largely to AI helping build better AI.
AP reported on Nov 5, 2026 that the Berkeley nonprofit Model Evaluation and Threat Research (METR) published an independent analysis of Anthropic. Anthropic gave METR unprecedented access to about 11 weeks of internal data (mid-August to late October 2026) plus staff interviews. METR estimated Anthropic is improving its models on well-specified tasks about twice as fast as a year earlier. It ruled out computing power as the main driver and believes the gain largely comes from AI helping to develop newer AI. METR said AI cannot yet make fully autonomous contributions to the most important parts of model training, though Anthropic researchers see 'signs of life'. It warned the figure could be overstated because gains may apply mainly to easily graded tasks, and redactions limited what it could publish. In a Sept 12 essay, CEO Dario Amodei urged the industry to slow capability growth so safety can keep up and endorsed embedded external evaluators; OpenAI's Sam Altman said he also supports them. Anthropic is preparing to go public later in November in a possibly record IPO. Anthropic did not immediately comment.
