Anthropic models accelerate gains as METR study points to AI-driven development

Traders Union
All News
Business & Wealth
Companies
Anthropic models accelerate gains

Highlights

Anthropic's benchmark capabilities have improved 1.9 times faster since October 2025, according to an 11-week METR assessment using confidential data.
METR estimates that AI contributes approximately 80% of the recent acceleration in model development, with no major bottlenecks expected in the coming year.
The prospect of recursive self-improvement raises concerns about existential risks and could reshape industry economics after hundreds of billions of dollars invested in infrastructure.

Study findings on model development

As reported by Reuters, citing METR, the AI safety nonprofit also known as Model Evaluation and Threat Research, Anthropic's benchmark capabilities have been improving 1.9 times faster since October 2025 than before, based on an 11-week assessment using confidential research, interviews and performance data from unreleased models.

The findings, published on Thursday, provide some of the clearest evidence yet that AI is helping speed up the creation of more capable systems. METR estimates AI accounts for perhaps 80% of the current acceleration, while cautioning that the figure remains uncertain.

The researchers say they find no convincing evidence of major bottlenecks likely to slow progress over the next year. METR also says it is confident the acceleration does not stem from faster growth in training compute scaling.

Implications for AI safety and industry investment

METR cautions that AI is not yet capable of making highly significant, fully autonomous contributions to the most important stages of model training. Even so, the results add to concern around recursive self-improvement, a scenario in which AI systems build more powerful versions of themselves without human intervention.

That prospect has fueled warnings from AI researchers in recent months that such systems could become harder to align with human goals, monitor and control, potentially creating existential risks. The shift could also reshape the economics of the sector, where companies have already spent hundreds of billions of dollars on computing infrastructure to train advanced models.

Our earlier coverage of international AI safety coordination highlighted how governments and leading developers are exploring stronger safeguards as AI capabilities advance rapidly. We noted Nvidia CEO Jensen Huang’s call for engagement with Chinese AI companies, discussion of a U.S.-China AI incident channel, and the idea of the UK as a neutral venue to help manage cross-border risks and prevent misuse.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.