Anthropic Calls for Worldwide AI Development Pause Amid Self-Improvement Concerns

The Lead: Anthropic's Global AI Pause Proposal

Anthropic has called for a worldwide "temporary pause" on AI development, announcing plans to convene policymakers to discuss the potential dangers of advanced AI systems. The proposal comes in the company's latest release detailing the progress of its AI model, Claude, toward "recursive self-improvement" capabilities that could enable it to create more powerful versions of itself.

The Recursive Self-Improvement Concern

Anthropic's Thursday post highlights a "trend" of increasing capability in Claude which, "taken far enough and given enough compute... points to an AI system capable of fully autonomously designing and developing its own successor." This development raises concerns about "humans losing control over AI systems." The concept of recursive self-improvement is viewed by AI safety researchers as a critical step toward superintelligence, potentially unleashing widespread consequences on humanity.

The company notes that as of May 2026, more than 80% of the code merged into Anthropic's codebase was authored by Claude, which is becoming increasingly proficient at "steering research" and "proposing its own experiments" within coding-related tasks.

The Government Partnership

The news comes alongside reports that Anthropic has embedded engineers inside the National Security Agency, despite a legal battle with the Pentagon over the use of its tools. These engineers are reportedly helping the NSA use Anthropic's model Mythos for offensive cybersecurity operations.

This dual approach—calling for global AI safety discussions while supporting military applications—has drawn criticism. Steven Murdoch, a professor at University College London, noted that "Anthropic might give the impression of being warm and fuzzy, but their definition of AI safety is narrow. Supporting US authorities in the development of offensive capabilities has never been something they have spoken against."

The Skepticism Response

Some experts question whether there's truly a new development prompting Anthropic's call for a pause. Murdoch stated that "Anthropic's post did not offer evidence of any step changes in the progress of AI capabilities" and suggested that "nothing has fundamentally changed today that has caused Anthropic to publish this article."

He also noted that Anthropic's call for a "temporary pause" on AI echoes other proposals on AI safety the company has made throughout the years. "It's a reminder of what they are concerned about, and have been concerned about for many years," Murdoch added.

The Future Outlook

Two months ago, Anthropic announced Mythos, an AI model they claimed was too powerful for the public due to cybersecurity concerns, though some experts called the announcement "a marketing post" with more hype than substance.

The company's latest moves come as Anthropic filed for an IPO that could value the company at $1 trillion, signaling the growing commercial and strategic importance of AI technologies in the global landscape.