AI will not necessarily lead to environmental ruin or proliferation, discrimination or equality, enhancement or overreliance - it can do each of those and more.

This the first of six laws considering the role of technology in society – set out by Melvin Kranzberg, President of the Society for the History of Technology, in 1985. His insight effectively captures the challenge of understanding the relationship between AI and our pursuit of sustainability objectives.

Technology is neither good nor bad: there’s no deterministic conclusion that we should draw from the trajectory that AI will follow, the way in which it will impact our sustainability goals, or how we pursue them.

Nor is it neutral: any technology is as impartial as its human user. If we don’t selectively apply its capabilities and consider the potential implications – AI can propagate a variety of unwanted outcomes or unintended consequences.

AI holds tremendous potential – but without due consideration, AI technologies can also threaten progress towards achieving some of the UN’s Sustainable Development Goals (SDGs).

‘Bigger is better!’ is the firmly touted belief among current leaders in AI development – reflected in the focus on increasing the size of models by raising the number of parameters, volume of training data, and computing power to improve performance. Scaling large language models (LLMs) has certainly proven an effective strategy over recent years.

Whether scaling is a necessary or sustainable route to further progress or leveraging AI’s potential remains an important and nuanced debate – but even the current scale of LLMs or foundation models is plying pressure on global energy supplies.

Implementation challenges include the need for high-quality data, technical expertise, and robust infrastructure. The responsible development of AI solutions requires careful attention to ethical considerations, privacy concerns, and inclusive design principles.

There’s also need for greater consideration of social aspects, with research indicating that less than 5% of AI projects for SDGs focus on goals like Gender Equality (SDG 5) and Clean Water and Sanitation (SDG 6).

Ethical dilemmas

If AI systems used to assess environmental impact are trained primarily on data from developed economies, they may fail to accurately represent conditions in emerging markets or underserved communities. Consequently, the risk of perpetuating existing inequalities through AI-driven sustainability initiatives requires counterbalancing with localised modelling and data collection.

Moreover, as organisations begin to leverage AI for sustainability reporting and decision-making, the ability to make a critical assessment of AI outputs is essential to avoid discriminatory actions. Transparency and accountability are, therefore, primary considerations.

The application of AI in sustainability reporting requires careful consideration of both algorithmic quality and ethical principles, particularly when distinguishing between mandated ESG disclosures and reputation-oriented sustainability communications.

This differentiation becomes especially crucial as organisations navigate complex reporting landscapes – where requirements can vary significantly between regions and standard-setters.

Not A Climate-Only Approach

Taking a climate-first, but not a climate-only approach, these standards demand a new level of sophistication in how organisations track, analyse and report their environmental impact. Organisations must now provide comprehensive insights into their climate-related risks – from immediate physical threats to long-term transition challenges.

Technology is supporting this evolution of reporting requirements in myriad ways. Much of sustainability reporting is fundamentally data-driven, yet many companies do not have sufficient data flows and structures in place to meet these requirements. So, tools to support materiality assessments and data-driven reporting are increasingly in-demand.

By efficiently collecting and analysing data from multiple sources, AI systems can help streamline the materiality assessment process – and provide more accurate risk and opportunity analysis.

Effectively leveraging AI for sustainability reporting and other objectives ultimately depends on user education and cultural change. Organisations need to focus on getting people to think about AI as something that’s learning from them – encouraging people to input and maintain data that will provide more value. At the heart of these challenges lies the fundamental issue of data quality and standardisation.

Organisations must also consider their data governance practices. Research suggests that up to 80% of on-premises business data may be ‘dark data’ – redundant or obsolete information that reduces AI training utility while increasing storage-related environmental costs