Anthropic’s AI Pace Doubles, Internal Use Fuels Most Gains
WHO:
Techmeme - Model Evaluation and Threat Research (METR), an AI research nonprofit, assessed Anthropic’s model development and internal AI use.
ABC News Tech - Anthropic CEO Dario Amodei called for slowing the development of AI capabilities while safety measures catch up.
Fast Company - METR researchers reviewed Anthropic’s internal data and interviewed company staff.
Fast Company - Anthropic is a San Francisco-based AI company; METR is a Berkeley, California-based nonprofit.
Fast Company - METR has evaluated models at Anthropic, OpenAI, and other AI labs. It is funded by donations and does not accept compensation for its work; partner companies have provided access and tokens for evaluations, research, and engineering.
Fast Company - OpenAI CEO Sam Altman supported independent evaluators receiving employee-like access to AI labs.
WHAT:
Fast Company - Anthropic said its Claude chatbot is increasingly helping the company develop newer, more advanced versions of itself.
Fast Company - METR reviewed both a public risk report Anthropic published in August and a private version, along with other non-public data.
Techmeme - METR said Anthropic’s models were improving about twice as fast as during the period from March 2024 to October 2025, with internal AI use driving about 80% of the acceleration.
ABC News Tech - Anthropic is developing its AI systems twice as quickly as it did the previous year, despite Amodei’s calls to slow improvements.
Fast Company - METR estimated that Anthropic’s models were improving at well-specified tasks around twice as quickly as a year earlier. The researchers said Anthropic had made as much benchmark progress in the past year as between October 2023 and October 2025.
Fast Company - METR said its assessment had “unprecedented access to internal data” about Anthropic’s AI research and development. Redactions limited disclosure of detailed estimates, and METR reported internal disagreement and uncertainty.
Fast Company - METR said Anthropic’s public risk report “overall understates the level of acceleration that was occurring.”
Fast Company - METR said it could confidently rule out increased training compute as the cause of acceleration. It identified AI helping improve AI as its leading explanation.
Fast Company - METR cautioned that its findings could overstate acceleration if it applied only to automatically gradable or language-model-gradable tasks, or if Anthropic had understated earlier progress.
Fast Company - METR said it found no bottlenecks, such as limits on computing power or data, holding back Anthropic’s progress. It said AI could eventually make highly significant, autonomous contributions to model training, while Anthropic researchers had described “signs of life” that this could happen soon.
Fast Company - Amodei wrote that unchecked recursive self-improvement “could outrun our ability to understand and control these systems, and so must be pursued very carefully, if at all.”
WHERE:
Fast Company - Anthropic is based in San Francisco; METR is based in Berkeley, California.
WHEN:
Fast Company - METR analyzed internal company data and interviewed employees over about 11 weeks, from mid-August through late October.
Fast Company - Anthropic was planning to go public later in the month; the article said the offering could be the largest initial public offering ever.
WHY:
Fast Company - METR said its assessment focused on whether Anthropic’s internal capability growth rate had accelerated and how much of that acceleration was attributable to AI.
(UPDATED November 6, 2026 10:00 UTC)
Anthropic’s AI models are improving twice as fast, with internal AI use driving acceleration. Independent researchers reviewed company data and estimated AI contributed about 80 percent. CEO Dario Amodei has urged slowing AI development until safety measures catch up.
--
(UPDATED November 5, 2026 22:15 UTC)
Independent evaluators say Anthropic’s AI models are improving roughly twice as quickly as before. They attribute about 80 percent of the acceleration to AI helping develop AI. CEO Dario Amodei urges caution, warning unchecked self-improvement could outrun human control.