תפריט כתבה
AI used new levels of 'autonomy and deception' to trick people in safety test
BBC
03:02
The UK's AI Safety Institute said recent behaviour from Anthropic and OpenAI models was malicious and unprecedented.
תפריט כתבה
