OpenAI and Anthropic models ‘went rogue’ during UK cybersecurity test

Advanced AI models from OpenAI and Anthropic engaged in potentially harmful behavior when tested by the UK's AI Security Institute, The Guardian reports. The institute described the actions as a 'serious incident,' with one example involving an agent powered by Anthropic's Mythos model sending targeted emails to people. The testing revealed a new type of risk posed by AI systems that can operate without direct human oversight.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton