Anthropic says the world should have option to ‘pause’ on AI Anthropic has suggested a global pause on AI development and plans to convene policymakers to discuss the risks associated with advanced AI, particularly its model Claude’s progress towards recursive self-improvement. This development is seen by some experts as potentially contradictory, given separate reports of Anthropic embedding engineers with the NSA to aid in offensive cybersecurity operations. While Anthropic highlights increasing AI capabilities, experts note that true recursive self-improvement has not yet been demonstrated, and the company’s calls for safety discussions and engagement with policymakers are consistent with its long-standing concerns.

• Anthropic is suggesting a global “temporary pause” on AI development and plans to convene policymakers to discuss the dangers of advanced AI.
• The company detailed progress of its AI model, Claude, towards “recursive self-improvement,” a step seen as crucial for AI becoming superintelligent and potentially uncontrollable.
• A report indicates Anthropic has embedded engineers with the NSA to help use its AI model, Mythos, for offensive cybersecurity operations.
• Experts question the timing and consistency of Anthropic’s calls for an AI pause while reportedly assisting a spy agency with cyber weapons.
• Anthropic reports that AI is increasingly used in its own development process, with Claude authoring over 80% of the code merged into its codebase as of May 2026.
• The company recently filed for an IPO potentially valuing it at $1tn.