New Study: Telling People You Used AI Might Backfire

Companies and governments are pushing for AI disclosure: telling people when content was made or helped by artificial intelligence. The idea sounds fair, like a label on a food package. But a new research paper asks an uncomfortable question: who actually pays the price for that honesty?

Researchers ran a workshop at a major tech fairness conference. They asked participants to design real-world scenarios around AI disclosure in schools, workplaces, politics, and personal relationships. What they found is that disclosure is not neutral. Admitting you used AI can make people see you as less competent, less trustworthy, or even lazy. One example: a worker who uses AI accessibility tools, like speech-to-text, might get punished in performance reviews for relying on them. The same tools that help them do their job become a mark against them.

This burden is not shared equally. People in already marginalized groups, such as disabled workers or minority professionals, are more likely to face suspicion and surveillance when they disclose AI use. Meanwhile, people with power and privilege often disclose AI use without consequence, or simply don't have to. The study calls this the "Cost-of-Honesty Stack" and suggests disclosure rules should be flexible and context-aware, not one-size-fits-all. The goal is transparency without turning honesty into a weapon against the vulnerable.