Artificial intelligence company, Anthropic has warned that AI systems may soon reach a stage where they can improve and develop themselves with minimal human involvement, prompting renewed calls for an industry-wide pause in the race to build increasingly powerful models.
In a widely discussed blog post authored by Jack Clark and Marina Favaro, the company said advances in its AI capabilities suggest the industry could be approaching a milestone known as “recursive self-improvement” — a process where AI systems autonomously design, build, and train more advanced versions of themselves.
According to the authors, AI tools are already playing a major role inside Anthropic. They revealed that more than 80 per cent of code merged into the company’s codebase is now generated by its AI assistant, Claude, while engineering teams are reportedly producing nearly eight times more code per quarter than they did before 2025.
Anthropic cautioned that although AI systems have not yet crossed the threshold into autonomous self-improvement, such a development could arrive sooner than many governments, businesses, and institutions are prepared for.
The company warned that failing to establish adequate safeguards before reaching that stage could make it increasingly difficult for humans to maintain meaningful oversight and control over advanced AI systems.
Call for Coordinated Global Pause
As part of its recommendations, Anthropic advocated for a temporary pause in frontier AI development, but only under conditions where multiple leading AI laboratories in different countries agree to halt development simultaneously.
The proposal would likely require cooperation among major industry players. However, analysts note that such cooperation appears unlikely given the intense competition among AI firms.
Timing Raises Questions
Anthropic’s warning comes at a pivotal moment for the company. The AI firm recently filed confidential paperwork for a potential initial public offering and has reportedly emerged as one of the world’s most valuable artificial intelligence companies..
Critics have previously accused Anthropic of using AI safety concerns as a competitive strategy aimed at slowing rivals, an allegation the company has consistently rejected.
The renewed call for caution follows changes to Anthropic’s own safety framework earlier this year.
The company previously maintained a policy that committed it to halting the training of advanced AI systems if adequate safety measures could not be guaranteed beforehand. However, that commitment was later revised.
Anthropic executives argued that unilateral restraint would have limited impact if competitors continued advancing their systems at full speed. Instead, the company adopted a framework focused on transparency and industry-wide safety standards.
Anthropic said it plans to engage governments, researchers, civil society groups, and industry stakeholders in broader discussions about the risks and governance challenges posed by advanced AI development.
