Scholarlink

<p>Can binding AI regulation reshape enterprise adoption before its substantive obligations fully apply? We theorise regulatory enactment as an anticipatory sociotechnical coordination signal. When the legal consequences of AI are perceived as ambiguous, organisations face an institutional information constraint: they cannot readily allocate responsibility, evaluate vendors or authorise wider deployment. A credible framework may relax that constraint enough to deepen existing adoption, even while first-time adoption remains limited by skills, data, budgets and infrastructure. We test this argument using the EU Artificial Intelligence Act and Eurostat enterprise ICT surveys for 2021, 2023, 2024 and 2025. A triple-difference design compares EU and non-EU country-sector cells with different pre-policy levels of respondent-reported legal ambiguity. With country-sector, country-year and sector-year fixed effects, a one-standard-deviation higher pre-policy ambiguity is associated with a 1.46 percentage-point increase in the share using at least two AI technologies and a 1.22 percentage-point increase in the share using at least three; the extensive margin does not increase significantly. The share of non-adopters reporting unclear legal consequences falls by 2.07 percentage points. Positive technology and functional margins are concentrated in text mining, speech recognition, ICT security and R&amp;D or innovation, whereas production-process adoption contracts. The central contribution is therefore a selective-coordination account of regulation: perceived regulatory clarification can change the depth and composition of adoption without producing broad diffusion. The estimates describe early country-sector responses, not individual-firm behaviour, long-run structural effects, firm-size heterogeneity or the disappearance of operational compliance uncertainty.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Perceived regulatory clarity and the depth of enterprise AI adoption: anticipatory evidence from the EU Artificial Intelligence Act

  • 期刊论文

  • 发表日期：

    2026年8月4日

查看全文
  • XiaoXi Ma,
  • Lu Chao,
  • Sun ShuiMiao,
  • Jing SaiChen

摘要

Can binding AI regulation reshape enterprise adoption before its substantive obligations fully apply? We theorise regulatory enactment as an anticipatory sociotechnical coordination signal. When the legal consequences of AI are perceived as ambiguous, organisations face an institutional information constraint: they cannot readily allocate responsibility, evaluate vendors or authorise wider deployment. A credible framework may relax that constraint enough to deepen existing adoption, even while first-time adoption remains limited by skills, data, budgets and infrastructure. We test this argument using the EU Artificial Intelligence Act and Eurostat enterprise ICT surveys for 2021, 2023, 2024 and 2025. A triple-difference design compares EU and non-EU country-sector cells with different pre-policy levels of respondent-reported legal ambiguity. With country-sector, country-year and sector-year fixed effects, a one-standard-deviation higher pre-policy ambiguity is associated with a 1.46 percentage-point increase in the share using at least two AI technologies and a 1.22 percentage-point increase in the share using at least three; the extensive margin does not increase significantly. The share of non-adopters reporting unclear legal consequences falls by 2.07 percentage points. Positive technology and functional margins are concentrated in text mining, speech recognition, ICT security and R&D or innovation, whereas production-process adoption contracts. The central contribution is therefore a selective-coordination account of regulation: perceived regulatory clarification can change the depth and composition of adoption without producing broad diffusion. The estimates describe early country-sector responses, not individual-firm behaviour, long-run structural effects, firm-size heterogeneity or the disappearance of operational compliance uncertainty.