Anthropic’s AI models are improving nearly twice as fast as they did before, according to an independent study that points to a growing role for the technology in building its own successors. The findings were published on Thursday by METR, an AI safety nonprofit, which said the warnings about such progress outpacing efforts to assess risks are now more pressing.
The report describes the results as some of the clearest evidence so far that AI is speeding up its own development. That trend could lead to recursive self-improvement, or RSI, in which AI systems build more capable versions of themselves without human involvement.
How METR measured the acceleration
The 11-week assessment, which ran from August to October, drew on confidential research, interviews and performance data from unreleased models. It found that Anthropic’s benchmark capabilities had been improving 1.9 times faster since October 2025 than before.
METR, also known as Model Evaluation and Threat Research, estimated that as of October 2026 Anthropic was likely advancing its models on well-specified tasks at roughly double the pace seen between March 2024 and October 2025. The reviewer stressed high uncertainty, saying it would not be surprised if the true acceleration was between 1.5 and 2.5 times the historical rate.
The group also said AI was not yet capable of making highly significant, fully autonomous contributions to the most important stages of model training.
Productivity claims and the planned public listing
Anthropic researchers have reported sharp productivity gains from using AI to write code and run experiments. Informal employee polls suggested improvements of about 12 times, but METR cautioned that those figures were unreliable.
Anthropic, whose chief executive called for slowing advanced AI development in September, did not immediately respond to a request for comment. The San Francisco-based company is preparing to go public, potentially as soon as this month, in what could be one of the largest IPOs ever.
Risks and implications for the wider industry
Researchers have warned in recent months that recursive self-improvement could make AI systems hard to align with human goals, monitor and control, potentially posing an existential threat to humanity.
Such a shift could reshape an industry that has spent hundreds of billions of dollars on computing infrastructure, since AI itself could become a major driver of technological progress. METR said it was confident the acceleration was not caused by faster growth in training compute.
