AICDI said most organizations are in the early stages of understanding both the operational and environmental impacts of AI.
The Thomson Reuters Foundation has released the first findings from the AI Corporate Data Initiative (AICDI). AICDI is the world's largest dataset to date examining how 1,000 companies across 13 sectors disclose and manage their use of artificial intelligence (AI).
The initiative, developed in partnership with UNESCO, reveals the widening gap between the rapid adoption of AI and the systems to manage its risks.
According to the report, companies are integrating AI at a much faster pace than they can establish safeguards. One of the most notable gaps is in environmental monitoring. 97% of companies did not assess the environmental impact of their AI systems, including energy consumption and carbon emissions.
AI is rewiring global finance
Evaluate the efficiency of AI tools
Despite frequent references to “ethical,” “trustworthy,” or “safe” AI, there is little linking the increased adoption of this technology to increased electricity demand, emissions, or companies' efforts to tackle climate change.
Energy efficiency experts like Excelgeo say organizations need to evaluate AI tools not only for governance and compliance risks, but also for their impact on energy performance, an increasingly important issue as AI workloads increase.
Optimizing AI-enabled systems can reduce energy demands, avoid unnecessary operation of boilers and chillers, and prevent simultaneous switch-on of equipment.
“Many people never ask about the environmental impact of AI because they think it always wastes energy, but that’s not true,” said Donatas Karczauskas, CEO of Exergeo.
“There are tools for AI to do the opposite: reduce consumption. For example, advanced building management systems use AI to reduce demand for heating and cooling, rather than increasing it.”
Make AI policies more accessible
The findings also revealed internal governance contradictions. While 76% of companies with an AI strategy report oversight at the management level, only 41% give employees access to AI policies or require employee approval, indicating poor adoption.
Karčiauskas said, “This study reveals a governance gap around measurement. If you're not monitoring AI behavior in real time, you're left guessing whether it's helping or hurting your goals.”
“In a building, you need to know when your systems turn on, how much power they consume, and what actually changes once the AI goes live. Without that operational data, AI governance is just red tape.”
EMEA region leads in publishing AI strategy
Regionally, companies in Europe, the Middle East, and Africa (EMEA) are leading the way in publishing their AI strategies, with 53% reporting an AI strategy. This trend is primarily driven by compliance with EU AI laws. However, even within Europe, environmental considerations, particularly energy use and emissions, are rarely included in AI governance disclosures.
Karčiauskas said: “Although regulation is progressing in Europe, the energy footprint of AI is still largely undiscussed here.
“As EU AI law matures, operational transparency, such as how much electricity the AI uses and whether it saves electricity, needs to be built in as part of governance. Otherwise, it is too easy to sell ‘responsible AI’ on paper, ignoring what real-world energy use looks like.”
AICDI concludes that most organizations are in the early stages of understanding both the operational and environmental impacts of AI.
The report highlights the need for companies to get a clearer picture of where AI is used within their companies, how it impacts performance, and what social, environmental and regulatory risks it may pose.
