EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

AI Washing or Substantive Adoption? A Game-Theoretic Model of Corporate AI Disclosure under Investor Scrutiny and Regulatory Enforcement

Chengcheng Wu, Zhiqiang Wang

Journal of Business and Data Analytics · 30.03.2026 · 4 (7) · 25-39

DOI: 10.63646/rwgj5521 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Business and Data Analytics, online erschienen am 30.03.2026. Verfasst von Chengcheng Wu und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Firms increasingly advertise their use of artificial intelligence (AI), and securities regulators have begun to sanction “AI washing”, that is, claims of AI capabilities that firms do not have. This study develops a game-theoretic model in which firms decide whether to adopt AI at a privately known cost and then choose between a substantive, verifiable AI claim and a vague one. Investors price firms on the basis of their disclosures, and a regulator audits substantive claims and sanctions false ones. Under the D1 refinement, the perfect Bayesian equilibrium is unique and moves through four regimes as the expected sanction rises: silence, pooling washing, partial washing, and truthful disclosure. Closed-form expressions are derived for the adoption rate, the share of firms making substantive claims, and the share of those claims that are false. Two results matter for business data analytics. First, in the partial-washing regime, measures that count substantive AI claims in corporate text overstate true adoption by a factor equal to the price stake of a claim divided by the expected sanction, and stronger enforcement raises true adoption while lowering measured claims. Second, the welfare-maximizing audit intensity is interior, has a closed form, and tolerates some washing; without commitment, the regulator does not audit and informative disclosure disappears. Numerical experiments, validated by an exhaustive equilibrium search and Monte Carlo simulation, show that the optimal policy raises adoption from 25% to 35% of firms and welfare by 7.0% relative to no enforcement, whereas full deterrence costs more than it gains. Sensitivity analysis and extensions to audit errors and sanction caps identify when these conclusions change.

Zitieren

APA

Wu, C., & Wang, Z. (2026). AI Washing or Substantive Adoption? A Game-Theoretic Model of Corporate AI Disclosure under Investor Scrutiny and Regulatory Enforcement. Journal of Business and Data Analytics, 4(7), 25-39. https://doi.org/10.63646/rwgj5521

Vancouver

Wu C, Wang Z. AI Washing or Substantive Adoption? A Game-Theoretic Model of Corporate AI Disclosure under Investor Scrutiny and Regulatory Enforcement. Journal of Business and Data Analytics. 2026;4(7):25-39. doi:10.63646/rwgj5521

BibTeX
@article{Wu2026,
  author = {Wu, Chengcheng and Wang, Zhiqiang},
  title = {AI Washing or Substantive Adoption? A Game-Theoretic Model of Corporate AI Disclosure under Investor Scrutiny and Regulatory Enforcement},
  journal = {Journal of Business and Data Analytics},
  year = {2026},
  volume = {4},
  number = {7},
  pages = {25--39},
  doi = {10.63646/rwgj5521}
}
Kostenlos starten →