Anthropic Proposes Coordinated Pause on Advanced AI Development

Anthropic, creator of the Claude chatbot, announced a proposal for leading AI firms to coordinate a temporary slowdown or pause of advanced AI development, warning that rapid progress could outpace human control.

NewserAI Billionaires Want to Control Every Aspect of Your Life | Aaron Bastani Meets Karen HaoIs AI voice control in cars the next big thing... or a complete waste of time?

Anthropic’s IPO is expected to be among the year’s largest, reflecting Claude’s popularity with developers. Since Claude Code launched in February 2025, the model has become central to software engineering, with the firm reporting that over 80 % of code published in May was generated by Claude, up from single-digit usage before its release.

Anthropic notes that AI models are accelerating, completing coding tasks orders of magnitude faster than earlier versions. The firm warns that an AI could design its own successor—a scenario termed “recursive self-improvement.”

Newser

In a blog post, co-founder Jack Clark and research director Marina Favaro argue that a coordinated pause would give time for alignment research and regulatory development. OpenAI’s report counters that democratic governments—not private labs—should set the rules, emphasizing that no single entity should dictate AI’s pace.

Newser

OpenAI argues that decisions about AI pace should not be left to any single lab, company, or special interest group, emphasizing that democratic governments must set the rules. University of Toronto researcher Nicolas Papernot demonstrated AI-driven “worm” attacks that adapt as they spread, highlighting security risks beyond the most powerful models.

Newser

Anthropic’s proposal lacks details on enforcement, verification, or safeguards against a “bad actor” bypassing a pause. OpenAI’s report similarly offers no concrete framework for governmental oversight, leaving both positions without actionable implementation plans.

Newser

• “it would be good for the world to have the option to slow or temporarily pause” — Jack Clark, co-founder, Anthropic

Newser

• “democratic governments—not private companies acting alone—must ultimately determine the rules, safeguards, and accountability mechanisms.” — OpenAI report

Newser

• “Our view is that decisions about the pace of AI innovation should not be left to any one lab, company, or special interest group,” — OpenAI report

Newser

• “It's not just the biggest, most powerful language models that pose the security concerns,” — Nicolas Papernot, lead researcher, University of Toronto

Newser

If AI achieves recursive self-improvement, it could outpace human oversight, raising safety concerns. Adaptive AI-driven cyber-weapons, such as “worms,” further heighten the need for coordinated governance, as unchecked acceleration may affect scientific, economic, and security domains.

Newser

Anthropic plans to convene its research institute with peer labs to design verification protocols, while policymakers worldwide are expected to further debate rules and safeguards for AI pace control.

Newser