Anthropic called for top AI labs to consider slowing frontier AI development as advanced models may approach recursive self-improvement.
