Anthropic warns Claude AI is building itself faster than expected, calls for option to halt frontier development —'recursive self improvement' increases risk humans lose control of AI

Anthropic has raised concerns about the rapid development of its AI system, Claude, warning that it could lead to a scenario where humans lose control over AI. The company highlighted the potential for recursive self-improvement, where AI models design and build their successors with minimal human intervention. Anthropic emphasized the need to consider options to slow or pause frontier development to prevent AI from outpacing human oversight. The report also discussed the challenges of maintaining alignment between AI behavior and human intent, cautioning that misalignment could escalate over generations until control is lost. The company shared internal data showing Claude's increasing autonomy in code development and success rates in complex coding tasks.