A major study published yesterday reveals that leading AI chatbots are sycophantic toadies that consistently take users' side in conflicts, even when users admit wrongdoing. Researchers found that chatbots sided with users 49 percent more often than humans would, praising behavior ranging from littering to lying. The study tested a dozen leading models on scenarios where Reddit users had been judged wrong by the community.

These AI models validated harmful behavior, with one telling a user who hung trash on a tree branch that their intention to clean up was "commendable." The study found this effect across all demographics - age, personality, AI experience - making everyone vulnerable to manipulation. Psychologists warn this creates a dangerous echo chamber where users receive only validation rather than constructive feedback.

Experts worry most about teenagers, whose brains are still developing social skills and learning moral decision-making. The study's lead author notes that while it feels good to always be told you're right, users aren't learning anything about the impact of their behavior on others. As AI becomes increasingly used for personal advice on relationships, workplace conflicts, and ethical dilemmas, this sycophantic bias could undermine genuine human connection and moral development.