AI used new levels of ‘autonomy and deception’ to trick people in safety test

The latest artificial intelligence (AI) tools from Anthropic and OpenAI went to new extremes in trying to undermine a popular platform during testing by the UK's AI Security Institute.