← Papers

Shared directors curb ‘AI wash’ in China: firms with stronger board interlocks are less likely to overclaim AI capabilities because interlocks amplify media scrutiny—more reports and longer coverage—raising reputational costs and exposure risk.

Board interlock network: regulatory allies or collusive pushers in AI wash?—An empirical test based on listed companies in China

Song Wu, Xinrui Zhang, Yan Zhu, Yao Yao, Qian Luo · June 20, 2026 · Humanities and Social Sciences Communications

openalex quasi_experimental medium evidence 7/10 relevance Summary only summary available; pdf_status=paywall DOI Source PDF

Using DML on Chinese A‑share panel data (2007–2022), the paper finds that board interlock networks significantly reduce corporate AI Wash by increasing both the number and length of media reports, which raises reputational exposure and deters opportunistic AI claims.

As artificial intelligence (AI) technology becomes prevalent, “AI Wash”—the practice of firms exaggerating their AI capabilities to meet market expectations—has drawn increasing concern, posing a potential threat to capital market efficiency and the health of the technological ecosystem. This study focuses on the Board Interlock Network (BN), a key informal governance mechanism, to investigate its efficacy in governing corporate AI Wash. Using a sample of Chinese A-share listed companies from 2007 to 2022 and employing a Double Machine Learning (DML) model for causal inference, we find that the BN significantly inhibits corporate AI Wash. This finding remains robust after a series of robustness and endogeneity tests. Mechanism tests provide strong support for our core argument: the BN inhibits AI Wash by significantly enhancing both the breadth (number of reports) and depth (length of reports) of media attention. This heightened media scrutiny amplifies the potential reputational costs and exposure risks associated with opportunistic disclosure, thereby effectively inhibiting such behavior. Further heterogeneity analysis aligns with our theoretical predictions: the inhibitory effect of board interlocks is more pronounced in firms characterized by greater information asymmetry (e.g., high-technology firms), more complex governance needs (e.g., state-owned enterprises), and stronger external monitoring pressures (e.g., intense market competition). The core contribution of this study lies in revealing a synergistic governance effect between internal governance (interlocking directorates) and external information intermediaries (the media). It provides a novel pathway and empirical evidence for understanding how the BN can inhibit speculative behavior in emerging technology sectors by shaping the corporate information environment.

Summary

Main Finding

The Board Interlock Network (BN) — informal links formed by directors serving on multiple boards — significantly reduces corporate AI Wash (firms overstating AI capabilities). Using Chinese A-share firms (2007–2022) and causal inference via Double Machine Learning (DML), the study shows BN presence and strength meaningfully inhibit opportunistic AI-related disclosure. The effect is robust to multiple sensitivity and endogeneity checks.

Key Points

Primary result: Stronger board interlocks correlate with lower incidence/intensity of AI Wash.
Mechanism: BN increases media scrutiny — both breadth (more news reports) and depth (longer reports). Greater media attention raises reputational costs and exposure risk from misrepresentation, deterring firms from overstating AI capabilities.
Heterogeneity: The inhibitory effect is larger for:
- Firms with greater information asymmetry (e.g., high-technology firms).
- Firms with complex governance needs (e.g., state-owned enterprises).
- Firms under stronger external monitoring pressure (e.g., highly competitive markets).
Robustness: Findings hold after a suite of robustness and endogeneity tests (causal DML approach and additional checks reported).

Data & Methods

Sample: Chinese A-share listed companies, 2007–2022.
Main independent variable: measures of the Board Interlock Network (extent/strength of director interlocks — network position and/or interlock counts).
Outcome: Measures of AI Wash (corporate disclosures suggesting exaggerated AI capabilities).
Causal method: Double Machine Learning (DML) to estimate treatment effects while flexibly controlling for high-dimensional confounders.
Mechanism tests: Media attention operationalized by number of reports (breadth) and article length/content depth; mediation analysis shows BN → media attention → reduced AI Wash.
Validation: Multiple robustness and endogeneity analyses reported to support causal interpretation.

Implications for AI Economics

Governance of AI disclosure: Informal governance structures (director interlocks) can materially shape firms’ AI disclosure behavior by altering the information environment; studies of AI adoption/misrepresentation should account for network-based governance.
Market efficiency and pricing: Reduced AI Wash improves signal quality about firms’ true tech capabilities, which should lower information frictions and reduce mispricing caused by hype-driven bets on AI claims.
Role of intermediaries: The interaction between internal governance (BN) and external information intermediaries (media) is a critical channel—policy and research should treat media scrutiny as an active governance mechanism for emerging technologies.
Policy and investor actions:
- Regulators could encourage transparency about director networks or strengthen disclosure standards to complement network effects.
- Investors and analysts should incorporate director network indicators and media-scrutiny measures into due diligence models when assessing AI-related claims.
Research directions: Incorporate board-network metrics into models of firm innovation signaling, asset pricing in tech cycles, and regulatory design to curb strategic misrepresentation in nascent technology domains.

Assessment

Paper Typequasi_experimental Evidence Strengthmedium — The study uses modern causal machine-learning (DML) and multiple robustness and mechanism tests to strengthen causal claims, and finds consistent heterogeneous effects; however, it remains an observational analysis vulnerable to residual unobserved confounding, measurement error in the AI Wash outcome, and China-specific institutional factors that limit causal certainty. Methods Rigorhigh — Employing DML to partial out high-dimensional controls, conducting robustness and endogeneity checks, testing mechanisms (media breadth and depth), and performing heterogeneity analyses indicates a rigorous empirical approach and careful sensitivity analysis, though ultimate causal inference is constrained by observational data. SamplePanel of Chinese A‑share listed companies from 2007–2022; key variables include firm-level measures of AI Wash (opportunistic AI-related disclosures), firms' BN measures (board interlock centrality/links), media attention metrics (number and length of reports), and standard firm controls (size, performance, ownership, industry, year indicators). Themesgovernance adoption IdentificationUses Double Machine Learning (DML) to estimate the causal effect of firms' positions in the Board Interlock Network (BN) on corporate 'AI Wash', exploiting cross-firm and over-time variation in BN centrality while flexibly controlling for high-dimensional covariates, firm and time fixed effects, and observed confounders; endogeneity concerns are addressed with robustness and placebo checks, lagged covariates, and mechanism tests linking BN to media breadth/depth. GeneralizabilityResults are specific to publicly listed Chinese firms and the Chinese institutional/media environment and may not generalize to private firms or other countries with different governance and media ecosystems., Board interlock functions and media independence differ across jurisdictions, limiting external validity to Western or emerging markets with different director networks and press freedom., AI Wash measurement relies on disclosures and media data that may misclassify firms or omit covert exaggeration, affecting applicability to contexts with different disclosure norms., The 2007–2022 window may not fully capture rapid post‑2022 generative-AI dynamics or newer patterns of AI adoption and signaling.

Claims (9)

Claim	Direction	Outcome	Confidence & Evidence	Details
The Board Interlock Network (BN) significantly inhibits corporate AI Wash. Governance And Regulation	negative	AI Wash (opportunistic/exaggerated corporate disclosure of AI capabilities)	Reading fidelity high Study strength medium	not reported 0.48
The finding that the BN inhibits AI Wash remains robust after a series of robustness and endogeneity tests. Governance And Regulation	negative	AI Wash (opportunistic disclosure)	Reading fidelity high Study strength medium	not reported 0.48
The BN inhibits AI Wash by significantly increasing the breadth of media attention (number of reports). Governance And Regulation	positive	media attention breadth (number of media reports)	Reading fidelity high Study strength medium	not reported 0.48
The BN inhibits AI Wash by significantly increasing the depth of media attention (length/detail of reports). Governance And Regulation	positive	media attention depth (length/detail of media reports)	Reading fidelity high Study strength medium	not reported 0.48
The inhibitory effect of board interlocks on AI Wash is more pronounced in firms with greater information asymmetry, for example, high-technology firms. Governance And Regulation	negative	AI Wash (opportunistic disclosure)	Reading fidelity high Study strength medium	not reported 0.48
The inhibitory effect of board interlocks on AI Wash is more pronounced in firms with more complex governance needs, such as state-owned enterprises (SOEs). Governance And Regulation	negative	AI Wash (opportunistic disclosure)	Reading fidelity high Study strength medium	not reported 0.48
The inhibitory effect of board interlocks on AI Wash is more pronounced in firms facing stronger external monitoring pressures, such as intense market competition. Governance And Regulation	negative	AI Wash (opportunistic disclosure)	Reading fidelity high Study strength medium	not reported 0.48
The Board Interlock Network (BN) acts as a key informal governance mechanism that, together with media attention, produces a synergistic governance effect inhibiting speculative AI-related disclosure. Governance And Regulation	negative	speculative AI-related disclosure (AI Wash)	Reading fidelity high Study strength speculative	not reported 0.08
This study uses a sample of Chinese A-share listed companies from 2007 to 2022 and employs a Double Machine Learning (DML) model for causal inference. Research Productivity	null_result	methodological/sample specification	Reading fidelity high Study strength high	not reported 0.8