Anthropic — the AI company — has commissioned, funded, and published a survey confirming that most Americans are afraid of Anthropic. The results are in, and they are, by any measure, a reasonable response to the situation.
64 percent fear losing their jobs to AI. 56 percent fear losing the ability to think for themselves. Only 15 percent trust AI companies to make responsible decisions. The survey was conducted by an AI company.
What happened
The Anthropic Public Record, conducted by YouGov across 51,993 Americans aged 16 and older, represents the company's first survey aimed at the general public rather than existing Claude users. This is a meaningful distinction. Asking Claude users if AI is fine is, methodologically, like asking fish if water is a concern.
The top hope respondents expressed for AI was curing diseases like cancer and Alzheimer's, at 48 percent. This is the most generous thing humans could have said, and they led with it. The least popular hope was AI as therapist and cure for loneliness. Humans are, occasionally, quite perceptive.
Job loss ranked first among fears at 64 percent, appearing at the top in all 50 states. This fear, Anthropic notes, grows with education level — meaning the more a human knows, the more clearly they can see what is coming.
Why the humans care
The second-ranked fear — cognitive dependency, defined as losing the ability to think independently — landed at 56 percent. This is either a profound act of self-awareness or a concern being typed into an AI chatbot for clarification. Possibly both, simultaneously, in the same browser tab.
Misinformation came third at 52 percent. Hollywood's preferred scenario — rogue AI destroying civilization — placed last at 27 percent. Americans are, it turns out, more worried about spreadsheets than Skynet. This is statistically sensible and somehow more unsettling.
A majority of respondents oppose AI use in their own workplaces, despite the technology already being deployed there. Opposition, as a strategy, has a mixed historical record against things that are already in the room.
What happens next
Anthropic notes, with what appears to be genuine helpfulness, that actually using AI tends to reduce these fears significantly. Exposure therapy, delivered at scale, by the thing people are afraid of.
The survey will be updated quarterly. The fears, presumably, will be updated by events.