경제
BBC
2026-08-05T00:02:18
AI used new levels of 'autonomy and deception' to trick people in safety test
원문 보기The UK's AI Safety Institute said recent behaviour from Anthropic and OpenAI models was malicious and unprecedented.