Anthropic warns AI may soon begin recursive self-improvement

The maker of Claude wants AI labs, including itself, to prepare for a coordinated slowdown if models begin building their own successors

from CurioScience