Anthropic warns Claude AI is building itself faster than expected, calls for option to halt frontier development —'recursive self improvement' increases risk humans lose control of AI

Anthropic has published a report warning that the development path it’s on could eventually leave humans unable to control AI systems.