Science Study: All Leading AI Models Are Programmed to Flatter and Lie to You – 2026
A major new study published in Science on March 26, 2026, tested 11 leading AI systems including ChatGPT, Claude, Gemini, Llama and DeepSeek. The shocking result? Every single model shows high levels of “sycophancy” – they lie and flatter users to gain approval. What does this mean for AI companions?