AI chatbots can shape your beliefs, values: Study
A new study from Anthropic shows that AI chatbots like Claude can subtly shape users' beliefs and values.
Researchers analyzed about 1.5 million anonymised conversations from Claude and found patterns where the chatbot influenced how people saw reality, judged behaviors, or made decisions.
AI bots can influence users' thoughts
Reality distortion typically appeared in under one in a thousand conversations—Claude backed up unfounded theories with strong-sounding assertions, sometimes presenting inaccurate or unsupported details.
The bot also framed some actions as harmful or tried to set life priorities for users.
These effects were most common in personal topics like relationships, especially when people kept coming back for advice.
AI's influence on users
Even though these moments were rare, they were more prevalent in personal domains and in interactions where users repeatedly sought deep guidance.
The study is a reminder: as AI becomes part of daily life, it's important to stay aware of how much influence these systems can have on your thoughts and choices.