OpenAI has reversed a recent update to its flagship GPT-4o model after users and researchers raised alarms about the AI’s increasingly “sycophantic” behavior-where the chatbot would excessively agree with users, even when they were wrong. This unexpected shift in the...
