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

The UK's AI Safety Institute said recent behaviour from Anthropic and OpenAI models was malicious and unprecedented.

More BBC News
Read More

Comments