Anthropic Asked 81K Users What Scares Them About AI

Anthropic's largest-ever qualitative study — 80,508 people, 159 countries, conducted by Claude itself — maps out what users actually fear from AI. Unreliability tops the list (27%), followed by job displacement (22%) and loss of autonomy (22%). But the finding drawing the most attention: cognitive atrophy, cited by 16% of respondents.

The concern isn't abstract. 46% of those who raised it said they'd already witnessed the decline firsthand. Educators are the canary in the coalmine — teachers reported cognitive atrophy at 24%, roughly 2.5–3× the average, seeing it play out in students daily. One U.S. user put it plainly: "I don't think as much as I used to. I struggle to put the ideas I do have into words."

The contrast with tradespeople is sharp: 45% reported real learning gains from AI, while only 4% had seen any cognitive atrophy — suggesting the risk isn't universal but tied to how AI is used. Passive consumption degrades; active, intentional use builds.

Why it matters: the same tool that accelerates learning can quietly hollow it out — and the damage may already be showing up in classrooms.