Where AI in Sustainability Can Go Wrong

What happens when AI works with weak sustainability data?
AI can produce outputs that appear precise even when the underlying information is incomplete, inconsistent or based on estimates and proxies. Understanding data lineage and making assumptions explicit are therefore essential when using AI for sustainability analysis.

How can a technology designed to support sustainability create environmental impacts?
AI systems require significant computing resources, with data centres consuming energy and water. Responsible use means considering the footprint of the technology itself, choosing the right-sized model for the task and avoiding unnecessary AI use where simpler approaches are sufficient.

Why does representativeness matter in sustainability applications?
AI outputs can reflect the limitations of the data used to develop models. Data-rich regions, populations or supply-chain tiers may be better represented than others. Users should question whose perspectives are represented, identify potential gaps and sense-check results against local and lived experience.

Why is overreliance a sustainability risk?
If professionals stop challenging AI-generated conclusions, important evaluation skills can gradually weaken. Human oversight remains particularly important for consequential work, where AI should support analysis rather than replace professional judgement.

Why can polished AI-generated content create problems?
AI can generate sustainability claims that sound credible without having a sufficiently strong evidence base behind them. A polished statement is not proof of accuracy. Important claims must be traceable to verifiable data and sources.

Do these risks mean organisations should avoid AI?
No. They provide a practical watch-list for using AI responsibly. The goal is neither to reject the technology nor to adopt it uncritically, but to develop the judgement and controls needed to capture its value while managing its limitations.