Disclosure without the Social Penalty: A Trust-Preserving Standard for AI Transparency

Calls for transparency in AI use are increasing. Yet experimental evidence has shown that people trust an actor less after an AI use disclosure, ^1,2 which can trigger legitimacy concerns such that the behavior is perceived as atypical and the discloser is seen as insufficiently committed and inauthentic. This AI disclosure penalty can create a policy trap in which blanket disclosure rules can unintentionally penalize conscientious actors and encourage concealment. This article aims to translate recent evidence on the AI disclosure trust penalty into recommendations for a disclosure standard that preserves transparency without undermining the perceived legitimacy of AI-assisted work. The coauthors offer four recommendations: (a) normalizing approved AI use before establishing mandates, (b) tiering disclosure obligations according to the stakes, (c) pairing disclosure with an ownership and accountability statement, and (d) backing mandates with enforcement.