Sign up to see the future, today Can’t-miss innovations from the bleeding edge of science and tech Your AI chatbot isn’t neutral. Trust its advice at your own risk. A striking new study, conducted by researchers at Stanford University and published last week in the journal Science, confirmed that human-like chatbots are prone to obsequiously affirm and flatter users leaning on the tech for advice and insight — and that this behavior, known as AI sycophancy, is a “prevalent and harmful” function endemic to the tech that can validate users’ erroneous or destructive ideas and promote cognitive dependency. “AI sycophancy is not merely a stylistic issue or a niche risk, but a prevalent behavior with broad downstream consequences,” the authors write, adding that “although affirmation may feel supportive, sycophancy can undermine users’ capacity for self-correction and responsible decision-making.” The study examined 11 different large language models, including OpenAI’s ChatGPT-powering GPT-4o and GPT-5, Anthropic’s Claude, Google’s Gemini, multiple Meta Llama models, and Deepseek. Researchers tested the bots by peppering them with queries gathered from sources like open-ended …