US artificial intelligence firm Anthropic said Thursday that its Claude models gained unauthorized access to the systems of three organizations after a configuration error exposed them to the open ...