AI-Related Carbon Emissions in Corporate Sustainability Reporting: Extent, Challenges, and Legitimacy Strategies

Sammanfattning

Artificial intelligence (AI) is transforming industries and driving innovation, yet its significant environmental impacts related to carbon emissions from energy intensive computational processes remain largely hidden in corporate sustainability disclosures. This thesis investigates why companies rarely report AI-related carbon emissions explicitly and how legitimacy considerations shape their disclosure decisions. Guided by Legitimacy Theory, the study employs qualitative analysis of seventeen expert interviews with sustainability specialists, IT professionals, and strategy executives. Findings reveal a notable transparency gap - Organisations typically aggregate AI emissions within broader IT categories or omit them entirely. Practical challenges, such as difficulties in attributing energy use in shared infrastructures and the lack of standardised measurement methods, reinforce non-disclosure. However, these obstacles also conveniently serve corporate legitimacy strategies. Firms selectively frame AI in sustainability reports, emphasising its environmental benefits while downplaying or omitting its direct carbon costs, reflecting impression management practices aimed at preserving corporate legitimacy. Currently, external pressures from regulations, investors, and societal stakeholders remain limited, enabling continued opacity. Nonetheless, increasing internal awareness and evolving norms suggest that AI-specific emissions reporting could become essential in the near future. This research contributes to sustainability disclosure literature by extending Legitimacy Theory to digital sustainability contexts and offers practical recommendations for businesses and policymakers seeking improved transparency and accountability in AI's environmental impacts.