AI NEEDS BETTER EMISSIONS DATA TO BECOME SUSTAINABLE

The rapid expansion of AI systems has raised concerns about their environmental footprint, but progress on sustainability remains stalled by fundamental gaps in data and understanding. Luccioni argues that the industry cannot address emissions problems without knowing the true environmental cost of training and running large language models. Current measurement standards are inconsistent, making it difficult to compare the efficiency of different systems or track improvements over time. Beyond emissions tracking, there's a secondary problem: limited visibility into how AI is actually being used. Without understanding real-world deployment patterns, optimizing for sustainability becomes guesswork. The researcher's position suggests two concrete requirements for progress. First, standardized methods for measuring and reporting AI emissions across the industry. Second, better data collection on AI usage patterns to identify where efficiency gains matter most. Without these foundational elements, sustainability claims remain largely unverifiable and industry incentives to reduce environmental impact stay weak.