Why responsible AI matters now

AI adoption in at least one business function rose to 88 per cent of companies in 2025, up from 78 per cent the year before (p.14). Yet only 13 per cent of S&P 500 companies have directors with AI expertise (p.15), and just 14 per cent of boards regularly discuss AI (p.15), signalling a widening governance gap.

The AI company Data Initiative

Launched in 2024, the AI Company Data Initiative (AICDI) is a free, voluntary framework developed by the Thomson Reuters Foundation in partnership with UNESCO. The dataset covers 2,972 companies across 11 sectors and five geographies, generating more than 100,000 data points to enable comparable benchmarking of responsible AI governance.

The regulatory and framework landscape

The EU AI Act is the world’s first comprehensive AI law, classifying systems by risk level and mandating requirements including human oversight and documentation. UNESCO’s Recommendation on the Ethics of AI — endorsed by 193 countries — and the OECD AI Principles offer non-binding normative guidance. Where binding laws are absent, corporate self-regulation remains a major force shaping governance practice.

Finding 1: Public commitment to AI governance frameworks remains low

Only 13 per cent of companies say they align their strategy with a formal AI governance framework (p.10), while 43.7 per cent publicly communicate having an AI strategy (p.10). Of those citing a framework, 53 per cent reference the EU AI Act (p.10), indicating that AI strategies are frequently developed without a corresponding external commitment to recognised governance frameworks.

Finding 2: Many companies publish strategies on AI but it is less clear how these are put into practice

40 per cent of companies report board- or committee-level oversight of AI (p.10), yet fewer than a third have a dedicated AI governance team (p.36). Only 2.7 per cent report having a formal AI model registry (p.37), and just 12.4 per cent have a policy to ensure a human oversees AI systems (p.10), suggesting governance is often described conceptually rather than operationally.

Finding 3: Companies do not demonstrate adequate protections for workers as AI reshapes jobs

While 31 per cent of companies claim to have AI training programmes, only 12 per cent offered structured training with comprehensive coverage (p.10). Only 14 per cent have policies to mitigate negative impacts of AI on workers (p.46), and just 2.3 per cent evidence an AI-related internal complaints mechanism (p.47). The report recommends companies set specific, measurable safeguards and actively engage workers in their development.

Finding 4: Ethical issues and environmental impacts are being sidelined

72 per cent of companies do not report conducting any impact assessment with regard to AI (p.10). Of those that do, only 7 per cent carry out Human Rights Impact Assessments and 11 per cent Environmental Impact Assessments (p.10). Data centres accounted for around 1.5 per cent of global electricity consumption in 2024, with demand projected to more than double by 2030 (p.55).

Finding 5: Limited company policies on AI training data, third-party data controls, and user data rights

Only 24 per cent of companies with AI strategies have policies to evaluate training data quality (p.57). Around one in five report policies for data sharing with third-party AI providers (p.59), and only 16 per cent allow users to request the deletion of their data and stop its processing by AI systems (p.61).

Sentiment analysis

Analysis of more than 15,000 open text fields reveals an overall tone of approximately 7 per cent positive, 92 per cent neutral, and 0.5 per cent negative (p.66). Human capital topics attracted the highest positive share, while governance and security themes remain more neutral, reflecting that many organisations are still building governance frameworks.

Guidance for investors

Investors face exposure across operational risk, regulatory headwinds, human capital vulnerabilities, and data governance gaps. AI-related proxy proposals attracted average support of about 30 per cent in the 2024–2025 proxy seasons (p.87). Investors are advised to focus engagement on meaningful transparency, board expertise, independent assurance, and concrete governance actions rather than high-level commitments.