Independent briefing Society & human agency
What 81,000 people want from AI
A large qualitative study of people’s hopes and fears, from professional excellence to time freedom and loss of control.
Independent briefing of Anthropic (2025). Read the original for its full evidence, context, and wording.
Start here
Three questions this article answers
What do people most often hope AI will provide?
Greater capability: better work, personal growth, easier life management, creativity, and more time for valued activities.
Are hopes and fears separate?
Often they are paired. The same person may want powerful assistance and worry about dependence, displacement, privacy, or loss of human meaning.
Whose views does the study represent?
A large, international sample of active Claude users—not a representative survey of the entire public.
Go deeper
Three related questions it doesn’t answer—matched to established papers
Can conversational AI steer users toward agreeable beliefs?
Experiments investigate how user feedback can reward agreeable rather than truthful model behavior.
Paper answering this questionSycophancy to SubterfugeDoes generative AI increase creativity while reducing collective diversity?
A preregistered experiment measures both individual story quality and similarity across outputs.
Paper answering this questionGenerative AI Enhances Individual Creativity but Reduces the Collective Diversity of Novel ContentHow should people shape the behavior of an AI system?
The study shows why users need a working mental model of an AI’s strengths and limits before relying on it.
Paper answering this questionHello AI: Uncovering the Onboarding Needs of Medical Practitioners for Human–AI Collaborative Decision-MakingThe useful summary
What to take away
This project uses tens of thousands of conversations to gather a richer picture than a multiple-choice poll. People want AI to help them become more capable, not merely more productive: to learn, create, manage life, and free time. But those goals sit beside concerns about employment, overreliance, privacy, and social disconnection.
The sample is unusually large and geographically broad, yet it is still selected: participants already use one company’s assistant and chose to engage. The results are best read as a deep map of active-user aspirations, not a vote from humanity. Product decisions should combine this kind of listening with representative research and special attention to people likely to bear harms.
Final conclusion
The three answers, brought together
People commonly want capability, growth, and time from AI, while fearing dependence and lost control. Those hopes and fears coexist. The study illuminates active Claude users across many countries, but it should not be mistaken for a representative global mandate.