Companies deploying AI ‘at unprecedented speed’ without considering ESG risks

A wide majority of companies deploying AI in their operations fail to consider environmental and societal impacts or set up sufficient governance structures, new analysis has found.

A review of 1,000 companies from 13 sectors has found that the corporate world lacks transparency around AI governance, potentially exposing investors to regulatory and reputational risks.

The Thomson Reuters Foundation’s AI Company Data Initiative (AICDI) found that 48% of the sample have disclosed AI strategies or guidelines suggesting they are implementing AI systems and tools “at unprecedented speed”.

Within the companies with an AI strategy, 71% included principles like ‘ethical’, ‘safe’, ‘secure’ or ‘trustworthy’ AI – yet the research revealed significant gaps: 97% of the overall sample did not consider environmental impacts such as energy use and carbon footprint, and 68% of those with AI strategies did not assess societal impacts beyond end users.

Additionally, more than three-quarters (76%) of companies with an AI strategy reported management-level oversight of AI, but only 41% made AI policies accessible to employees or required their acknowledgement, potentially creating a gap between policy and practice.

At a time when little is known about the full impact of AI, companies are exposing themselves and their investors to unknown risks by adopting the technology without proper risk assessments or governance, argues AICDI. 

To mitigate risks, the initiative recommends that investors start requesting an AI governance brief during due diligence, incorporating specific disclosures on oversight, transparency (including environmental and risk disclosures), and regulatory exposure, as well as benchmarking responses against sector and regional peers to identify leaders and laggards.

American organisations less stringent than in EMEA

Despite the fact that the US dominates AI innovation and adoption, the study found that American companies are less transparent than those in EMEA: just 38% published an AI policy, compared to over half (53%) in EMEA, for example.

Within the group of companies with AI strategies, 46% had dedicated AI governance teams in the EMEA region (which the report associates with the EU AI Act, which sets minimum governance obligations for AI deployers), versus only 26% in the Americas.

Certain sectors do better than others when it comes to AI governance and transparency: financials, communication services and information technology firms were three times more likely to have responsible AI teams or roles than those in the energy and materials sector.

Fossil fuel firms deploying AI to maximise output

While most firms may miss environmental risks in their AI implementation, oil and gas companies are also deploying the technology to maximise production, despite the well-known climate impacts of further fossil fuel extraction.

“Every day I see smart people talking about AI and climate, yet they’re missing an incredibly consequential piece: advanced technologies are actively expanding fossil fuel production, today, at scale, with no public accountability,” warns Holly Alpine, a former Microsoft employee who raised the alarm on the company’s contracts with fossil fuels producers.

“I honestly don’t know how to express how serious this is – or how alarming it is that it’s not even generally part of the debate, or the funding, around ‘AI’s impact on climate change’. If the largest oil producer on Earth doubles its productivity, it overwhelms the emissions reductions the rest of the world is trying to achieve,” she added, referring to recent statements by Saudi Aramco’s CEO on AI doubling oil well productivity.