AI labs are moving closer to systems that help improve their successors.

The Associated Press examined the push toward recursive self-improvement, a loosely defined idea in which AI systems help build stronger successors. Anthropic said Claude now contributes to 26% of its model research and development, but the work remains under human supervision rather than an autonomous improvement loop.

The report said OpenAI is targeting an automated AI researcher by March 2028 while acknowledging that it does not yet know how to safely reach aligned full recursive self-improvement. Elon Musk has said xAI is aiming for recursive self-improvement potentially by the end of 2026 and no later than 2027.