Analysis has shown that 97% of companies have failed to assess the environmental or energy impact of their AI systems, despite rapid adoption and growing regulatory focus on “responsible AI”.
This is based on research from the Thomson Reuters Foundation’s AI Company Data Initiative (AICDI), analysing AI governance disclosures from 1,000 companies worldwide.
The associated report indicates that this measurement gap matters most in energy-intensive sectors like commercial buildings, and how optimisation-focused AI can actually reduce energy use rather than increase it.
The accumulated data is said to be the world’s largest dataset on how firms describe and oversee their use of artificial intelligence, and the results reveal a wide gap between AI adoption and its governance – companies are deploying AI far faster than they are managing its risks.
The AICDI output also identifies a critical blind spot on the environmental side: 97% of companies did not assess the environmental impact of their AI systems, including their energy use or carbon footprint.
While many firms describe their AI as “ethical”, “trustworthy”, or “secure”, almost none connect AI deployment with rising electricity demand, emissions, or climate commitments.
Commenting on the findings, Donatas Karčiauskas, CEO of Exergio says: “Many people assume AI will always waste energy, so they never stop to ask about its environmental impact. But that’s not true. There are tools where AI does the opposite – it cuts consumption. Advanced building management systems, for example, use AI to lower heating and cooling demand instead of raising it”.
Karčiauskas is recommending that companies should evaluate which AI tools they adopt – not only for governance risks but also for how those tools affect energy performance.
Further, Karčiauskas notes that this gap is especially visible in sectors that operate large physical assets, such as commercial buildings, where poorly governed AI can lead not only to compliance risks but also to significant, unmeasured energy waste.
Karčiauskas has found that while 76% of companies with an AI strategy say AI is overseen at the management level, only 41% make their AI policies accessible to employees or require them to acknowledge those policies.
According to Karčiauskas, this is a clear warning sign that we evaluate AI’s impact wrongly, even on a regulatory level: “The study exposes a governance gap around measurement. If you don’t watch what AI is doing in real time, you’re guessing whether it helps or harms your goals. In buildings, that means knowing when systems switch on, how much power they pull, and what actually changes once AI starts running them. Without that operational data, AI governance is just paperwork”.
The analyst observes that this impact is most visible when AI is integrated into building management systems. Once established, such an approach can smooth demand peaks, keep boilers and chillers from running longer than needed, and prevent everything from switching on at the same time.
According to AICDI, companies in EMEA (Europe, the Middle East, and Africa) lead in publishing AI strategies, with 53% reporting one – largely driven by the EU AI Act. Yet even in Europe, the environmental impacts of AI, including energy use and emissions, are still rarely included in AI governance disclosures.
This leads Karčiauskas to comment: “Europe is ahead on regulation, but even here the energy footprint of AI is mostly absent from the discussion. As the EU AI Act matures, operational transparency – including how much power AI uses and whether it saves any – needs to be part of governance. Otherwise, it’s too easy to sell ‘responsible AI’ on paper while ignoring what happens to real-world energy use”.
