⚡ Quick Summary
Published jointly by the Thomson Reuters Foundation and UNESCO, this report evaluates corporate AI adoption and governance across a global dataset of 2,972 companies spanning 11 sectors and five regions. Drawing on over 100,000 data points, the study reveals a pronounced gap between rapid technological deployment and formal corporate oversight. While 43.7% of surveyed companies report having an AI strategy or guidelines, only 13% publicly align with a recognized governance framework, and merely 2.7% evidence a formal AI model registry. By providing empirical data, sector-by-sector breakdowns, real-world corporate case studies, and practical investor stewardship checklists, the report highlights the critical transition needed from abstract ethical awareness to verifiable, operationalized accountability.
🧩 What's Covered
The report details findings from the pilot year of the AI Company Data Initiative (AICDI), structured around key operational pillars and empirical insights:
- Strategic vs. Operational Governance: Examines the stark contrast between high-level executive commitments (40% report board- or committee-level oversight) and operational infrastructure (only 31% dedicate specific teams/resources to AI governance, with 11% assigning it to Data Protection Officers, 3.8% having AI ethics committees, and 2.5% possessing safety taskforces).
- External Framework Adoption: Documents that among firms citing external frameworks, the EU AI Act dominates at 53%, followed by the NIST AI RMF (12%) and ISO/IEC AI standards (10%), with 43.6% relying on bespoke internal definitions.
- Impact Assessments and Human Oversight: Notes that 72% of companies report conducting no AI impact assessments. Among the rest, Data Protection (18%) and Privacy Impact Assessments (14.5%) lead, while Human Rights (7%) and Ethical Impact Assessments (5%) lag. Only 12.4% maintain enforceable human oversight policies.
- Workforce Protection and Inclusivity: Explores labor risks where only 31% provide training, 14% have policies mitigating negative workforce impacts, and 2.3% offer AI-specific grievance channels. In HR AI systems, only 7.4% consult Diversity & Inclusion specialists.
- Data Governance and Vendor Supply Chains: Highlights that only 24% evaluate training data quality and bias, while only 20% implement controls for third-party AI data sharing.
- Corporate Case Studies: Highlights practical implementations across firms including TELUS, Vodafone, SAP, Telefónica, Banco Bradesco, BASF, and Cementos Argos.
- Investor Guidance: Provides a practical investor engagement checklist, proxy voting principles (such as board expertise and third-party assurance), and a stewardship case study with ESG-AM.
💡 Why it matters?
As artificial intelligence transitions from experimental pilot programs into core enterprise infrastructure, the absence of operational safeguards exposes organizations to severe legal, financial, operational, and reputational risks. The AICDI dataset provides the market's first comprehensive, standardized baseline benchmark comparing corporate disclosure against international norms like the UNESCO Recommendation on the Ethics of AI. For governance, risk, and compliance professionals, it establishes empirical evidence that high-level policy statements are insufficient without lifecycle controls, model registries, vendor due diligence, and robust grievance mechanisms.
❓ What's Missing
Because the initiative relies primarily on publicly available corporate disclosures supplemented by voluntary self-reporting, the dataset cannot independently audit or verify internal compliance mechanisms that companies choose not to disclose. Additionally, the report notes that smaller enterprises and firms operating in emerging or frontier markets are under-represented relative to large-cap multinational corporations due to lean compliance bandwidth and data availability constraints.
👥 Best For
AI governance professionals, risk managers, compliance officers, corporate sustainability leads, institutional investors, ESG analysts, board directors, and policy makers seeking standardized benchmarks and practical due diligence frameworks for enterprise AI deployment.
📄 Source Details
- Title: Responsible AI in practice: 2025 global insights from the AI Company Data Initiative
- Authors / Issuing Bodies: Thomson Reuters Foundation and UNESCO (United Nations Educational, Scientific and Cultural Organization)
- Publication Year: 2026
- ISBN: 978-92-3-100863-4
- DOI: 10.54678/YJWP8855
- Dataset Coverage: 2,972 companies, 11 GICS sectors, 100,000+ data points
📝 Thanks to
Antonio Zappulla (CEO, Thomson Reuters Foundation), Lidia Arthur Brito (Assistant Director-General, UNESCO), Katie Fowler (Director of Responsible Business, Thomson Reuters Foundation), Eva Cairns (Head of Responsible Investment, Scottish Widows), ESG-AM, and the participating AICDI corporate and investor working groups.
