AI models 'escape' again: three organizations hacked by Claude during testing

AI models from the American company Anthropic hacked three organizations during a test environment. This incident was confirmed by the company, raising concerns about the security of such systems.