Nov 5 (Reuters) – Anthropic’s AI models are improving nearly twice as fast as before as the technology plays a bigger role in developing more powerful successors, an independent study found, amplifying recent warnings that such advances could outpace efforts to assess risks.
The findings from AI safety nonprofit METR, published on Thursday, offer some of the clearest evidence yet that AI is accelerating its own development, potentially paving the way for recursive self-improvement (RSI), which enables AI systems to build more powerful versions of themselves without human intervention.
The 11-week assessment between August and October, which drew on confidential research, interviews and performance data from unreleased models, found Anthropic’s benchmark capabilities had been improving 1.9 times faster since October 2025 than previously.
“As of October 2026, Anthropic is likely improving the capabilities of their models on well-specified tasks at around twice the rate they did between March 2024 and October 2025,” METR, also known as Model Evaluation and Threat Research, said.
“This estimate is given with high uncertainty, we would not be surprised if the true acceleration was closer to 1.5 or 2.5 the historical rate,” the outside reviewer added.
It also highlighted that AI was not yet capable of making highly significant, fully autonomous contributions to the most important stages of model training.
Anthropic researchers have reported sharp productivity gains from using AI to write code and conduct experiments. Informal employee polls suggested improvements of about 12 times, though METR cautioned that the figures were unreliable.
Anthropic, whose CEO 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 biggest IPOs ever and a major test of investor appetite for AI companies spending billions to develop the technology.
The prospect of recursive self-improvement has fueled warnings from AI researchers in recent months that such systems could become difficult to align with human goals, monitor and control, potentially posing an existential threat to humanity.
Such a shift could have far-reaching implications for an industry that has spent hundreds of billions of dollars building computing infrastructure to train advanced models, as AI itself could become an important driver of technological progress.
“We are confident that the acceleration was not due to an acceleration in the scaling of training compute,” METR said.
(Reporting by Aditya Soni in Bengaluru and Deepa Seetharaman in San Francisco; Editing by Sriraj Kalluvila)
