Potential for Harm and Privacy Are Top AI Ethics Concerns: Evidence From a Representative UK Survey

Artificial intelligence is increasingly used in high-stakes decision-making, yet public acceptance depends on how ethical principles are weighed in specific contexts. We surveyed a representative sample of 300 UK adults to examine judgments of AI ethics in real-world applications. Participants rated the importance of fairness, transparency, privacy, nonmaleficence, and accountability, and evaluated six AI applications. All principles were strongly endorsed, with nonmaleficence and privacy rated highest. Perceived ethicality varied by context: healthcare diagnosis was judged most ethical, whereas autonomous weapons were judged least ethical. Women rated healthcare AI as less ethical than men. Endorsement of abstract ethical principles did not predict judgments in thematically matched scenarios, suggesting that case evaluations integrate multiple concerns. Positive general attitudes toward AI predicted higher ethicality ratings across domains, although this association weakened for autonomous weapons. Trust in automation, risk-taking, AI literacy, and moral foundations showed limited or no reliable effects.