Anthropic warns of AI self-improvement and suggests development pause (eco.sapo.pt)

• Anthropic warns AI could autonomously create its own successor soon.
• Over 80% of Anthropic's integrated code is now AI-generated by Claude.
• Company proposes a verifiable international pause on frontier AI development.

"Anthropic warned that artificial intelligence will be able to autonomously create its own successor within the next few years, suggesting a pause in development to allow society and research to keep pace. Over 80% of code incorporated by Anthropic is now written by its Claude model, up from under 10% in February 2025."