2025人工智能环境影响评估的关键维度与标准化路径研究报告

Disclaimers The designations employed and the presentation of material in this publication do not imply theexpression of any opinion whatsoever on the part of the International Telecommunication Union(ITU) or of the ITU secretariat concerning the legal status of any country, territory, city, or area or ofits authorities, or concerning the delimitation of its frontiers or boundaries. The mention of specific companies or of certain manufacturers’ products does not imply that theyare endorsed or recommended by ITU in preference to others of a similar nature that are notmentioned. Errors and omissions excepted; the names of proprietary products are distinguishedby initial capital letters. All reasonable precautions have been taken by ITU to verify the information contained in thispublication. However, the published material is being distributed without warranty of any kind,either expressed or implied. The responsibility for the interpretation and use of the material lieswith the reader. The opinions, findings and conclusions expressed in this publication do not necessarily reflect theviews of ITU or its membership. ISBN 978-92-61-40801-5 (PDF version)978-92-61-40811-4 (EPUB version) Please consider the environment before printing this report. © ITU 2025 Some rights reserved. This work is licensed to the public through a Creative Commons Attribution-Non-Commercial-Share Alike 3.0 IGO license (CC BY-NC-SA 3.0 IGO). Under the terms of this licence, you may copy, redistribute and adapt the work for non-commercial purposes,provided the work is appropriately cited. In any use of this work, there should be no suggestion that ITUendorse any specific organization, products or services. The unauthorized use of the ITU names or logos isnot permitted. If you adapt the work, then you must license your work under the same or equivalent CreativeCommons licence. If you create a translation of this work, you should add the following disclaimer alongwith the suggested citation: “This translation was not created by the International Telecommunication Union(ITU). ITU is not responsible for the content or accuracy of this translation. The original English edition shallbe the binding and authentic edition”. For more information, please visit https://creativecommons.org/licenses/by-nc-sa/3.0/igo/ Measuring what matters:How to assess AI’senvironmental impact Foreword by the ITU DeputySecretary-General As artificial intelligence (AI) transformsindustries, economies, and societies,itsrapid growth brings significantenvironmental implications. Energy consumption, water usage, andgreenhouse gas emissions associatedwith AI are rising sharply. Electricity consumption by data centresincreased 12 per cent each year from2017 to 2023, four times faster thanglobal electricity growth, according to theIEA. As per theGreening Digital Companies 2025report, four leading AI-focused companiessaw their operational emissions increase by 150 per cent on average between 2020 and 2023. But tech innovation must support, not hinder, climate action. Digital companies can drive decarbonization. Their technologies offer the potential to boostefficiencies across different industries, unlock innovative climate solutions, and addressnumerous socio-economic challenges. To achieve this, the global tech industry needs to better understand and manage positiveand negative impacts. The urgency of doing so has never been greater. Yet the potential forgreening digital remains obscured by inconsistent measurement practices and fragmentedaccountability. That is why ITU and partners worldwide, coming together forGreen Digital Action, have calledon the tech industry to take responsibility. With an array of governments, businesses, civil society and international organizations, we aimto establish common, transparent measurement criteria and reporting frameworks, promoteand uphold green technology standards, and build a circular economy. The Declaration on Green Digital Action, put forward by the Azerbaijan COP29 presidencyand ITU, has received endorsements from over 80 countries and nearly 1,800 companies andother non-state organizations worldwide. It calls for action to reduce environmental and climateimpact while ensuring the benefits of connectivity extend to everyone worldwide. As we approach the next climate talks at COP30, we expect new national and corporatecommitments to fully capture digital impact on climate – both negative and positive. This report,Measuring what matters: How to assess AI’s environmental impact, reviews commonassessment methodologies over the entire AI lifecycle. It stems from the Green Computing pillarof our Green Digital Action initiative, with a key sub-group on sustainable AI. We are examining implications at each phase of AI development, from initial models andtraining to real-world usages and interactions. By identifying gaps and flagging inconsistencies in current AI impact measurement, we aim toprovide actionable insi