Buyers' disclosure and use of AI-driven environmental oversight is linked to fewer environmental scandals at foreign suppliers, particularly where countries have stronger AI capacity and regulatory institutions.

Main Finding

Suppliers’ exposure to buyers’ AI-enabled environmental governance is associated with fewer supplier environmental controversies in the following year. This negative relationship is stronger when suppliers operate in countries with higher AI readiness and better regulatory quality.

Key Points

• Theoretical framing: combines Organizational Information Processing Theory (OIPT) and Signaling Theory.
  • OIPT: AI increases buyers’ capacity to collect, integrate, and process fragmented environmental information across borders, improving detection and oversight.
  • Signaling: buyers’ AI-enabled governance disclosures credibly signal heightened scrutiny, increasing suppliers’ perceived detection risk and deterrence.
• Empirical finding: greater exposure to buyer AI-enabled environmental governance → reduced supplier controversies next year.
• Moderators:
  • Supplier-country AI readiness amplifies the effect (better digital/data infrastructure improves cross-border information flows and interoperability).
  • Supplier-country regulatory quality amplifies the effect (stronger institutions increase credibility/enforceability and perceived oversight).
• Measurement innovation: buyer AI-enabled environmental governance measured via sentence-level text analysis of SEC 10-K filings; aggregated to supplier exposure.
• Sample: unbalanced panel of 2,505 foreign suppliers of non-financial U.S.-listed firms across 41 countries, 2020–2024.
• Estimation: multidimensional fixed-effects models with firm-clustered standard errors; numerous robustness checks (alternative measures, samples, estimation approaches, endogeneity mitigation).

Data & Methods

• Data sources: SEC 10-K filings (text), FactSet Supply Chain Relationships (buyer–supplier links), RepRisk (controversies), Oxford Insights (AI readiness), Worldwide Governance Indicators (WGI) and other firm- and country-level databases.
• Key variables:
  • Buyer AI-enabled environmental governance: sentence-level NLP of 10-K disclosures capturing firm-wide AI use for environmental information processing; aggregated to supplier exposure via observed buyer–supplier links.
  • Supplier environmental controversies: measured from reputational/controversy dataset (RepRisk or similar).
  • Moderators: country-level AI readiness (Oxford Insights) and regulatory quality (WGI).
• Empirical strategy:
  • Panel regressions with multidimensional fixed effects (controls for time, buyer, supplier dimensions as appropriate).
  • Standard errors clustered at firm level.
  • Robustness: alternative operationalizations, sample restrictions, and methods to address endogeneity (authors report comprehensive checks).

Implications for AI Economics

• AI as a governance technology with cross-border externalities:
  • AI adoption by buyers generates positive spillovers that can reduce negative externalities (supplier environmental misconduct) across supply chains.
  • AI governance can alter incentive structures beyond the adopting firm, affecting upstream behavior and risk allocation.
• Complementarity between firm-level AI and country-level infrastructure/regulation:
  • Returns to AI-enabled governance are higher where digital infrastructure (AI readiness) and institutional quality support information flows and enforcement—highlighting complementarities that matter for diffusion and social returns.
• Market and strategic effects:
  • Buyers that adopt and disclose AI governance may gain upstream risk-reduction benefits and reputational signaling advantages, potentially creating competitive pressure for other buyers and altering supplier selection dynamics.
  • Suppliers in low-AI-readiness / low-regulatory-quality countries may become disadvantaged (or remain higher-risk), which can affect global sourcing patterns and inequality across countries.
• Policy relevance:
  • Public investment in AI infrastructure and improvements in regulatory quality can magnify private-sector AI governance benefits for sustainability.
  • Standardization and interoperability of data/AI systems across jurisdictions would increase the effectiveness of private AI governance.
• Methodological contribution for AI economics research:
  • Demonstrates an NLP-based, disclosure-derived measure of AI-enabled governance suitable for large-sample empirical work; useful for future causal and structural analyses of AI’s economic impacts.
• Cautions and open questions:
  • Disclosure vs. capability: text-based measures may capture signaling as well as actual technical capacity—disentangling these is important for causal interpretation.
  • Residual endogeneity: authors perform robustness checks, but reverse causality or omitted variables may remain; causal mechanisms (detection vs. deterrence) warrant further identification.
  • Welfare and distributional effects: need quantification of net social benefits, costs of AI governance adoption, and cross-country distributional impacts.
  • Heterogeneity to explore: industry differences, supplier dependence on particular buyers, and potential unintended consequences (e.g., regulatory arbitrage, surveillance concerns).

Summary judgement: The paper provides credible evidence that buyer-side AI-enabled environmental governance can reduce supplier controversies and that country-level digital and institutional contexts matter. It advances understanding of AI’s role in cross-border governance spillovers and supplies a scalable disclosure-based measure useful for AI economics research, while leaving causal mechanisms and welfare implications for future work.