What concerns people about AI?

A new study from research firm ClearerThinking.org, published on LessWrong, provides a data-driven snapshot of American anxieties about artificial intelligence. The October 2025 survey of 403 US adults asked participants to rate their concern across 16 distinct AI-related risks, from immediate societal impacts to long-term existential threats. The findings reveal a population deeply worried about near-term harms: AI-generated misinformation and deepfakes topped the list, with 71% of respondents expressing concern, closely followed by AI-driven job elimination (70%) and the use of AI for scams and manipulation (69%).

The research uncovered significant demographic divides in AI anxiety. Women reported higher levels of concern than men across most categories, and self-identified progressives were more worried than conservatives. Perhaps most counterintuitive was the finding related to knowledge: participants who rated themselves as more knowledgeable about AI tended to be *more* concerned, not less, suggesting that familiarity breeds caution rather than comfort. Other high-ranking concerns included the proliferation of low-quality 'AI slop' (68%), AI bias and discrimination (67%), and the concentration of power in the hands of a few entities controlling advanced AI (66%).

Longer-term, speculative risks like superintelligent AI (57%) and 'slaughterbots' or weaponized drones (56%) still garnered majority concern but ranked lower than more immediate, tangible issues. The study methodology involved showing participants either short definitions or longer descriptions of each concern; interestingly, the length of explanation did not significantly alter worry levels. This data offers crucial insight for policymakers and tech developers about which public fears need addressing most urgently to build responsible trust in AI.