Anthropic warns about fully recursive self-improvement in AI: 'Humans may lose control'

Anthropic's research suggests AI is increasingly used to develop and improve new AI systems, with Claude now writing over 80% ...