Anthropic Warns Claude May Soon Build Itself

Anthropic has revealed that Claude now writes over 80 percent of its own code, up from less than 10 percent in early 2025. The AI firm warns this trend points toward recursive self-improvement, where AI could design its own successors without human intervention. While Anthropic acknowledges potential benefits in science and healthcare, it warns humans could lose control over AI systems. The company is calling for a coordinated slowdown requiring agreement among major labs across multiple countries, admitting the world lacks the time to build the necessary verification infrastructure.