Embedded Emissions Frameworks have long been a crucial tool for policymakers to account for emissions across borders and supply chains. Now, they're turning their attention to AI.
Artificial intelligence (AI) is usually discussed in climate policy circles through the lens of two familiar questions: how do we reduce AI’s environmental footprint, and how do we use AI to accelerate decarbonization? Both matter.
Today though, a third question is becoming unavoidable: how does AI change the way we measure the emissions embedded in the goods and services that move across borders? The answer to this question sits at the intersection of digital governance, international trade and environmental accountability.
Embedded Emissions Frameworks and AI's environmental footprint
To understand how we measure AI's cross-border emissions, it is crucial to understand Embedded Emissions Frameworks (EEFs).
EEFs are emerging as the backbone of climate-related trade policy. They underpin mechanisms like the EU’s Carbon Border Adjustment Mechanism (CBAM) and Corporate Sustainability Reporting Directive (CSRD), product‑level emissions disclosure rules, and supply‑chain due‑diligence regulations. Their purpose is simple: make the emissions inside traded products visible, comparable and governable. But AI is complicating that task. It is transforming production systems, reshaping supply chains and altering the nature of products themselves. At the same time, the complexity of implementing EEFs is growing so quickly that AI may be the only tool capable of managing it.
AI's footprint beyond data centres
The environmental footprint of AI is often discussed when it comes to the electricity consumption of data centres or the carbon intensity of training large models. That is certainly a factor – but energy consumption captures only part of the picture.
Upstream, digitalization creates environmental impacts across raw-material extraction, manufacturing, energy and water use and electronic waste. Semiconductor fabrication is resource‑intensive; chip manufacturing relies on mineral extraction and water‑heavy industrial processes; and the global hardware supply chain is carbon‑rich long before a single model is trained.
Downstream, AI influences logistics, product design, resource extraction and consumer behaviour. It can reduce emissions through optimization or amplify them by accelerating production and consumption.
These upstream and downstream effects raise the question that conventional AI sustainability metrics do not always answer: where, and how, should the environmental footprint of AI-enabled economic activity be accounted for? This is where EEFs become important.
The emerging EEF literature emphasizes the importance of accuracy, transparency, relevance and consistency, while also recognizing the need for methodologies that can operate within international trade rules.
Consider a product whose design, manufacturing process, logistics or energy management has been substantially optimized by AI. The environmental outcome may depend not only on the physical product but on the digital systems and infrastructure supporting its production.
Similarly, an AI-enabled service may rely on computing infrastructure located across multiple jurisdictions, powered by electricity with different emissions intensities.
This does not mean that every environmental impact associated with AI should automatically be incorporated into an EEF. Rather, it raises a more fundamental question about system boundaries and attribution: which environmental impacts are material to the good or service being traded, and what information is needed to account for them credibly?
How AI governance is addressing embedded emissions
The OECD AI Principles call for AI to contribute to sustainable development and environmental sustainability, while UNESCO's Recommendation on the Ethics of Artificial Intelligence explicitly calls for environmental and ecosystem impacts to be considered across the AI lifecycle, including energy consumption, carbon emissions and resource use. The environmental footprint of the AI ecosystem and value chain should also be considered. The EU AI Act also moves in this direction by requiring providers of general-purpose AI models to document known or estimated energy consumption. The European Commission is developing further approaches to measuring energy consumption and emissions associated with AI models and systems.
These developments introduce concepts that are highly relevant to EEFs: transparency, traceability, measurement and accountability.
If AI materially affects the environmental performance of goods or services, information about that impact may increasingly become part of the broader environmental information architecture through which markets and regulators assess those products.
AI could empower Embedded Emissions Frameworks
The relationship, however, is not one-directional. AI not only creates new challenges for embedded emissions accounting, but could also help solve some of them.
As climate-related trade policies proliferate, businesses increasingly have to navigate multiple emissions methodologies, reporting requirements, verification processes and supply-chain data requirements.
AI could help automate emissions reporting, reconcile datasets, map supply chains, identify missing information and detect anomalies. Regulators could use AI to support verification, identify inconsistencies and assess trade-related emissions risks. This is particularly important for smaller firms that lack the resources to maintain sophisticated emissions-accounting systems.
But there is an important qualification: AI cannot solve weaknesses in the underlying accounting methodology.
If emissions factors are unreliable, supply-chain data is incomplete or system boundaries are inconsistent, automating the process does not make the resulting information more credible and can exacerbate this governance problem.
The challenge is therefore not simply to use AI for EEFs, but to ensure that AI-enabled accounting evolves to be transparent, auditable and consistent with agreed principles.
The future will likely involve co‑evolution.
AI-specific environmental standards may increasingly require reporting of energy consumption, emissions and resource use across the AI lifecycle and supply chain. Those data could, where relevant and material, become inputs into broader environmental accounting and disclosure systems.
At the same time, AI could become part of the infrastructure used to implement EEFs by helping businesses manage increasingly complex data and helping regulators verify environmental claims. AI use here should also be carefully managed given its environmental impact.
This is already becoming visible in international policy discussions. The WTO's World Trade Report 2025 highlighted how AI is transforming the production and trade of goods and services and emphasises the importance of international cooperation as AI reshapes global trade.
AI and EEFs cannot be governed in isolation
AI is reshaping production systems and supply chains; EEFs are reshaping how emissions are measured and governed across borders. AI governance determines how digital systems are made accountable, while trade and climate policies shape how environmental information affects market access and domestic regulation.
The next generation of climate governance will therefore not only regulate emissions. It will increasingly govern the digital systems that measure, allocate, verify and influence those emissions.
AI's environmental footprint must be made visible. But AI can also help make environmental accountability workable.
The opportunity is to design EEFs that capture the environmental realities of an increasingly AI-enabled economy and AI systems that help make those accounting frameworks practical, credible and scaleable across global trade and compatible with climate policies.
It is important to align incentives, break silos and work with like-minded actors. We need to think creatively to explore more incentives to drive EEFs and AI systems in this direction, ranging from government procurement opportunities to tax reliefs. This requires dialogue and joint work among AI, environmental, trade and other communities regarding existing and future mechanisms such as CBAM and CSRD.
