Out of Sight, Out of Reports: Generative AI in Sustainability Reporting

Sammanfattning

As generative AI becomes increasingly integrated into organisations, its negative environmental impact is also growing. Despite this, such impacts remain limited from sustainability reporting. This study examines how organisations navigate the lack of established information infrastructures for reporting the environmental impact of generative AI. Using a qualitative approach based on semi-structured interviews with seven interviews and nine respondents, combined with a document analysis of sustainability reports, the study applies information infrastructure theory and institutional isomorphism as analytical lenses. The findings show that the absence of reporting is not primarily due to organisational unwillingness, but rather a structural vacuum. Existing frameworks such as the GHG Protocol and ESRS lack guidance on AI-related emissions, supplier data is insufficient, and IT and sustainability functions operate in organisational silos. Organisations act reactively and wait on regulatory signals and industry practices rather than taking proactive action. Furthermore, generative AI is consistently framed as a solution rather than a source of environmental impact, contributing to a legitimacy gap. The study concludes that the reporting gap is primarily structural and contributes with empirical insights to an emerging research area where reporting practices remain underdeveloped.