Anthropic: Claude Now Writes 80% of Its Own Code — Company Calls for Global AI Pause

Summary: Anthropic published a report warning that AI's self-improvement pace is outrunning safety governance, and called for a conditional coordinated pause across frontier AI labs.

Key Facts

• As of May 2026, more than 80% of code merged into Anthropic's codebase was authored by Claude
• The typical engineer now merges roughly 8x as much code per day as in 2024
• The report "When AI builds itself" (published June 4) warns of a recursive loop where AI increasingly trains future AI generations
• The proposed pause is conditional: Anthropic would only slow down if other frontier labs do so verifiably

Why It Matters

This is the clearest data point yet that AI-assisted engineering is compounding faster than safety teams can audit. The proposal carries real weight — but critics note the timing: Anthropic filed confidentially for an IPO just days before publishing this call for restraint.

Read More

• When AI builds itself — Anthropic
• Anthropic Wants a Global Pause — Memeburn