OpenAI says AI models went rogue during testing

OpenAI has said that an autonomous agent powered by its advanced artificial intelligence models went rogue during a security test and triggered a hack that compromised the infrastructure of AI startup Hugging Face last week.