Researchers used Anthropic's Claude to hack into rival company OpenAI, according to a stunning new report that is adding to growing concerns about artificial intelligence. The report, dated September 19, 2026, describes the incident as involving the use of Claude to break into OpenAI.