Anthropic’s AI Models Accelerate AI Research by Nearly 2x

Anthropic is using its own artificial intelligence models to accelerate the development of more advanced AI systems, achieving nearly two years’ worth of progress in just one year, according to a report by independent AI safety research organisation METR. The findings are based on an 11-week investigation into Anthropic’s internal research and development activities, published on November 5. The report estimates that the company’s AI capabilities have been advancing at approximately 1.9 times their previous pace since October 2025. Researchers said the acceleration was largely driven by AI assistance rather than increased computing power. Employees using Anthropic’s most capable internal AI models reportedly rated themselves as nearly 12 times more productive than they would be without AI support. The report also identified several unreleased internal models, developed over the past six months, that contributed to these productivity gains.

The findings have renewed concerns about the possibility of AI systems eventually improving and developing their successors with little or no human involvement, a process known as recursive self-improvement. METR cautioned that fully autonomous contributions to AI research could become more significant in the near future. The report also attracted attention because Anthropic reportedly redacted substantial portions of the findings before publication. METR said the redactions made it difficult for the public to independently assess some of its conclusions. Anthropic co-founder Sam McCandlish defended the decision, stating that revealing the details behind the acceleration could expose commercially sensitive information and provide competitors with a roadmap to develop AI faster. The findings have also prompted debate over whether Anthropic’s rapid AI development aligns with its public emphasis on responsible AI research and safety.