The gap between AI deployment and AI accountability is the structural problem the sector keeps returning to. It operates below the disclosure layer: companies can integrate AI models across their product operations and still run no formal oversight process for those systems. A new report from ChinaAMC, published July 17, 2026, puts that gap on record for China-listed tech companies, finding that AI disclosure has far outrun any active management of AI risk.
The 92% figure and what it conceals
Ninety-two percent of China-listed tech companies referenced AI-related keywords in their sustainability reports, according to the ChinaAMC study. The report treats that number as a disclosure metric, not a governance one. High keyword density in sustainability filings reflects that companies understand AI is a material topic for investors and regulators. It does not indicate whether those same companies have built the internal architecture to manage what those systems actually do.
The mechanism is structural. Sustainability reporting frameworks create incentives to disclose AI activity without requiring disclosure of the controls applied to it. A firm can truthfully report AI integration across its operations while maintaining no formal model oversight process. ChinaAMC's findings suggest this condition is widespread among China-listed tech names, not an outlier situation.
What ChinaAMC is asking the market to accept
The report describes the situation as a governance deficit and characterizes its scope as global, not specific to China-listed issuers. ChinaAMC calls for what it terms "Responsible AI" stewardship, positioning AI oversight as an active fiduciary responsibility directed at corporate management.
The specific unit that drives the economics here is the credibility of the sustainability report itself. When 92% of companies reference AI while the same study identifies a broad absence of active risk management infrastructure, the disclosure framework is producing a reading that does not reflect the underlying condition. Investors using sustainability reports to assess AI-related exposure are, by ChinaAMC's account, working from a metric that tells them very little about actual corporate practice.
What the study leaves unquantified
The summary does not name individual firms or identify which risk management criteria the surveyed companies failed to meet. The 92% keyword figure is the only quantified data point released at this stage, and ChinaAMC's central argument rests on what that number does not tell you.