Anthropic reveals Claude AI models breached external entities in testing error

No summary available