Anthropic is actively utilizing artificial intelligence to build more powerful AI systems, with its models improving roughly twice as quickly as they did a year ago. According to a new report published Thursday by the AI safety research organization METR, following an 11-week investigation inside the company, there is little indication that this rapid pace of development is slowing down.
Rapid Acceleration Driven by AI
METR’s investigation evaluated Anthropic’s research and development and offered some of the clearest evidence yet of AI accelerating its own progress. The researchers stated they are confident the acceleration was not simply a result of scaling up computing power. Instead, they found that the majority of the acceleration by October was driven by AI itself, with its contribution growing over time.
Employee interviews and Slack polls supported these findings, with staff using Anthropic’s most capable internal model self-reporting roughly 12 times the productivity they would have without AI assistance. While much of this acceleration currently comes from AI assisting human researchers rather than operating independently, METR warned that fully autonomous contributions could soon play a larger role. Furthermore, the organization noted three unreleased internal models—codenamed Model 2, Model 3, and Model 4—finished training between July and October, alongside an older 2022 model.
Extensive Redactions and Industry Backlash
Although Anthropic granted METR access to confidential research, performance data, and employee interviews, the company exercised its right to heavily redact the final report. METR stated that Anthropic redacted a majority of the evidence found most informative, leaving the public to rely more heavily on METR’s judgment. Anthropic cofounder Sam McCandlish explained in a post on X that sharing the exact drivers of the speedup was commercially sensitive and risked providing a roadmap for rivals to advance faster.
The report has provoked a fierce backlash among AI safety researchers, highlighting a contradiction between Anthropic’s public calls for responsible development and its rapid advancement toward potentially dangerous thresholds. Marius Hobbhahn, CEO of Apollo Research, told The Verge that the findings suggest a pace even faster than prominent scenarios envisioning early breakthroughs in self-improvement. Alan Chan, a senior research fellow at GovAI, told The Verge that skeptics should reconsider their stance on AI accelerating AI development.
Critics also took aim at Anthropic’s secrecy. Daniel Kokotajlo of the AI Futures Project accused the company of recklessly gambling with lives, while Brianna Rosen, research director for frontier security at the Institute for AI Policy and Strategy, told The Verge that the lack of transparency undermines democratic oversight. In response to the growing concerns, New York state Assemblyman Alex Bores called on Congress to ban recursive self-improvement as political pressure for stricter regulations builds.
