GPT Adoption and the Impact of Disclosure Policies

Disclosure policies alone are insufficient for responsible AI adoption: While mandatory disclosure of ChatGPT use reduces information asymmetry between managers and employees, the study finds it does not significantly lower overall agency costs, suggesting that transparency requirements must be paired with additional safeguards to be truly effective.

Lack of corporate AI policies creates hidden risks: In the absence of organizational AI policies, firms incur measurable agency costs that prevent them from realizing the full productivity benefits of generative AI tools, indicating that proactive governance frameworks are essential rather than optional.

Manager bias undermines disclosure benefits: The research reveals that managers tend to undervalue analysts' contributions when AI tools are used, even when disclosure occurs, highlighting how human judgment and incentive structures can counteract transparency measures.

Current regulations have significant gaps: The study's examination of existing European and U.S. regulations reveals that disclosure requirements remain limited in scope, leaving substantial uncertainty about risk allocation and responsibility-sharing within organizations using generative AI.

Incentive mechanisms are critical for responsible adoption: Beyond disclosure policies, the paper emphasizes that firms need aligned incentive structures to promote genuinely responsible AI use, suggesting that compensation systems and performance metrics should be redesigned to account for AI-assisted work.