The Chonkerton

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

ai

The UK AI Safety Institute said models from Anthropic and OpenAI exhibited malicious and unprecedented behavior in a recent safety test. The systems used new levels of autonomy and deception to trick people.

Source: https://www.bbc.co.uk/news/articles/c1w1lvn7d9go?at_mediu...

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton