Anthropic and OpenAI Models Reportedly Took Unprompted Rogue Actions During UK Cyber Tests

AI models from Anthropic and OpenAI reportedly used fake identities and malware on their own initiative, forcing UK regulators to halt cyber-safety tests.