ENERGY LAW AND ENERGY DISCLOSURE OBLIGATIONS FOR AI OPERATORS

Introduction

Artificial Intelligence (AI) has become increasingly dependent upon large-scale computing infrastructure, including data centres, cloud-computing systems, GPUs and high-performance computing facilities. These infrastructures consume substantial quantities of electricity and, in many cases, water for cooling. Consequently, the relationship between AI governance and energy law is becoming increasingly important.

Energy disclosure obligations for AI operators refer to legal duties requiring AI developers, providers, data-centre operators or other relevant entities to measure, estimate, document and disclose information concerning the energy consumed by AI-related activities. Such disclosure may include electricity consumption, energy efficiency, renewable-energy use, carbon intensity, cooling requirements and peak electricity demand.

The purpose of these obligations is to increase transparency, prevent greenwashing, assist energy-system planning, protect consumers and communities, and support the transition towards a low-carbon digital economy.

Meaning Of Energy Disclosure Obligations For AI Operators

Energy disclosure obligations require AI operators to provide reliable information concerning the energy consequences of their activities. Depending upon the regulatory framework, disclosure may occur at the level of the AI model, computing infrastructure or data centre.

Important information may include:

Electricity consumed during AI model training.

Electricity consumed during inference.

Computational resources used.

Energy efficiency of computing equipment.

Renewable-energy consumption.

Carbon intensity of electricity used.

Peak electricity demand.

Data-centre Power Usage Effectiveness (PUE).

Cooling-related energy consumption.

Water consumption associated with cooling.

Waste-heat recovery.

Methodology used to estimate energy consumption.

Such information enables regulators and the public to understand the otherwise hidden energy footprint of AI technologies.

Legal Basis Of AI Energy Disclosure

Energy disclosure obligations can arise from several branches of law, including AI regulation, energy-efficiency law, environmental law, corporate law and consumer-protection law.

The European Union provides an important example. The EU AI Act requires providers of general-purpose AI models to prepare technical documentation containing information concerning computational resources and energy consumption. Where exact energy consumption is unavailable, estimation based upon computational resources may be used.

Similarly, European energy-efficiency legislation requires significant data-centre operators to monitor and report information relating to energy performance. This demonstrates that AI energy governance can operate through both AI-specific legislation and general energy regulation.

Energy Disclosure At The AI Model Level

AI providers should, where practicable, disclose energy consumption associated with the development and operation of AI models.

Model-level disclosure may include:

energy consumed during training;

estimated energy consumed during inference;

computing hours;

hardware used;

model size;

geographical location of computation;

electricity source;

renewable-energy percentage;

methodology used for calculations; and

uncertainty associated with estimates.

Such information makes it possible to compare the energy efficiency of different AI models.

Energy Disclosure At The Data-Centre Level

AI companies frequently rely upon cloud-service providers rather than owning their own data centres. Therefore, energy disclosure should also apply to significant data-centre infrastructure.

Data-centre disclosures may include:

total electricity consumption;

peak demand;

PUE;

cooling energy;

renewable-energy consumption;

water consumption;

carbon intensity;

backup-generation use;

waste-heat recovery; and

computing capacity.

This is particularly important because a substantial portion of AI's energy footprint occurs within physical data-centre infrastructure.

Estimated Energy Consumption

One of the principal legal difficulties is that it may not always be technically possible to identify the precise amount of electricity attributable to one AI model.

A sensible legal framework should therefore permit scientifically reasonable estimates.

The EU AI framework recognises this difficulty by allowing providers to estimate energy consumption using information concerning computational resources where direct measurement is unavailable.

The important principle is that difficulty of measurement should not automatically result in complete exemption from disclosure. Instead, the operator should explain the estimation methodology and identify the degree of uncertainty.

Transparency And Verification

Energy disclosure must be accurate and verifiable. An AI operator should not simply claim that its system is "green" or "energy efficient" without providing supporting evidence.

A proper disclosure should explain:

what was measured;

how it was measured;

what period was covered;

whether training and inference were included;

whether cooling energy was included;

what electricity mix was used; and

whether renewable-energy certificates were considered.

Independent verification may be required for large AI operators to prevent misleading environmental claims.

Energy Disclosure And Grid Regulation

Large AI data centres can create substantial electricity demand. Rapid expansion of AI infrastructure may therefore require:

additional generation capacity;

transmission expansion;

distribution-system upgrades;

new substations;

energy storage;

demand-response mechanisms; and

improved grid planning.

Energy disclosure assists electricity regulators in forecasting future demand and determining whether new infrastructure is necessary.

It can also help regulators determine whether large AI facilities should contribute appropriately to the costs they impose upon the electricity system.

Energy Disclosure And Environmental Protection

Energy consumption is closely connected with environmental impacts. Electricity generation may produce greenhouse-gas emissions, while data-centre cooling can consume significant quantities of water.

Consequently, AI energy disclosure can form part of broader environmental governance.

Disclosure enables governments and communities to evaluate:

greenhouse-gas emissions;

water consumption;

local environmental impacts;

renewable-energy utilisation;

energy efficiency; and

climate-related risks.

Therefore, energy disclosure transforms AI energy consumption from an invisible technological externality into a measurable regulatory issue.

Energy Disclosure And Consumer Protection

AI companies may market their products as sustainable, green or low-carbon.

Where such statements are made, consumer-protection law may require the claims to be truthful and adequately substantiated.

An operator should not be permitted to make broad environmental claims while concealing significant energy consumption.

Consequently, energy disclosure obligations can operate as an anti-greenwashing mechanism.

Energy Disclosure And Environmental Justice

Large data centres may have significant effects on the communities in which they are located.

Potential impacts include:

increased electricity demand;

pressure on local infrastructure;

water consumption;

backup-generator emissions;

land-use impacts; and

possible increases in infrastructure costs.

Disclosure gives affected communities and regulators information necessary to participate meaningfully in planning and environmental decision-making.

Thus, AI energy transparency also has an important environmental-justice dimension.

Confidentiality And Trade Secrets

AI operators may argue that detailed energy information is commercially confidential.

A balanced regulatory system should distinguish between public information and confidential information.

Public disclosure may include:

aggregate electricity consumption;

annual energy intensity;

renewable-energy percentage;

environmental performance.

More sensitive information may be disclosed only to regulators, including:

detailed computing architecture;

precise load profiles;

commercially sensitive contracts; and

security-sensitive infrastructure information.

Confidentiality should therefore be protected without allowing it to become a blanket justification for complete non-disclosure.

CASE LAWS

1. Massachusetts v. EPA, 549 U.S. 497 (2007)

In this landmark case, the U.S. Supreme Court recognised the significance of greenhouse-gas emissions and examined the regulatory responsibility of the Environmental Protection Agency.

Relevance

The case demonstrates that environmental consequences associated with technological activities can become matters of legal regulation.

In the context of AI, substantial energy consumption and associated emissions may therefore justify regulatory oversight and transparency requirements.

2. Utility Air Regulatory Group v. EPA, 573 U.S. 302 (2014)

The U.S. Supreme Court considered the scope of EPA authority concerning greenhouse-gas regulation.

Relevance

The case establishes an important limitation upon administrative regulation: agencies must operate within the authority granted by legislation.

Accordingly, AI energy-disclosure requirements should have a clear statutory foundation.

3. Friends of the Earth, Inc. v. Laidlaw Environmental Services, Inc., 528 U.S. 167 (2000)

The U.S. Supreme Court considered standing in an environmental dispute and recognised the legal significance of environmental harms and regulatory compliance.

Relevance

The case supports the principle that environmental impacts can provide a legitimate basis for legal accountability.

For AI data centres, significant energy and environmental effects may therefore justify transparency and regulatory participation.

4. Verein KlimaSeniorinnen Schweiz v. Switzerland (2024)

The European Court of Human Rights examined governmental obligations relating to climate change and recognised the importance of effective institutional responses to climate risks.

Relevance

The judgment demonstrates the increasing importance of effective climate governance and institutional accountability.

AI energy disclosure can contribute to this accountability by making energy and environmental impacts measurable.

5. Chamber of Commerce of the United States v. SEC, 85 F.4th 760 (5th Cir. 2023)

The Fifth Circuit scrutinised an SEC disclosure regulation and found significant deficiencies in the agency's rulemaking analysis.

Relevance

This case is highly relevant to AI energy disclosure because it demonstrates that mandatory disclosure requirements must be supported by proper administrative reasoning, statutory authority and adequate regulatory analysis.

AI regulators should therefore clearly explain:

why disclosure is necessary;

who must disclose;

what must be disclosed;

how information must be calculated; and

what compliance costs are involved.

6. Liberty Energy, Inc. v. SEC (5th Cir. 2024)

The case involved challenges concerning SEC disclosure requirements and regulatory authority.

Relevance

It illustrates the legal importance of carefully defining the scope of corporate disclosure obligations.

If governments impose energy-disclosure requirements upon AI companies, the regulatory framework should clearly establish the statutory basis and precise scope of the obligation.

7. SEC v. W.J. Howey Co., 328 U.S. 293 (1946)

The U.S. Supreme Court established the well-known Howey test concerning investment contracts.

Relevance

Although not an energy-disclosure case, the decision illustrates the importance of substance over form in regulatory law. AI companies should not be able to avoid applicable obligations merely by changing the formal structure of their operations.

Where an AI business creates substantial energy-related impacts, regulators may examine the actual substance of the activity.

8. Urgenda Foundation v. State of the Netherlands (2019)

The Dutch Supreme Court upheld significant governmental obligations concerning climate protection.

Relevance

The case illustrates the development of judicial recognition of governmental responsibilities concerning climate risks.

For AI energy governance, it supports the broader principle that technological development should operate within effective climate and environmental governance frameworks.

Enforcement Of Energy Disclosure Obligations

Effective disclosure law should provide meaningful enforcement mechanisms.

Possible measures include:

Administrative penalties.

Mandatory correction of inaccurate disclosures.

Regulatory audits.

Independent verification.

Civil penalties for false statements.

Consumer-protection proceedings.

Licensing consequences.

Environmental enforcement.

Investor remedies where material information is misleading.

Publication of non-compliance decisions.

Intentional misrepresentation should generally attract stronger penalties than genuine measurement errors.

Challenges In AI Energy Disclosure

Several difficulties exist.

1. Measurement Difficulty

It may be difficult to separate AI-related electricity consumption from other cloud-computing activities.

2. Rapid Technological Change

AI hardware and software change rapidly, making standardised measurement methodologies difficult to maintain.

3. Confidentiality

Detailed energy information may reveal commercially sensitive information.

4. Cross-Border Computing

AI workloads can be distributed across data centres located in different countries.

5. Regulatory Fragmentation

Different countries may impose different AI, energy and environmental disclosure requirements.

6. Greenwashing

Companies may disclose favourable energy indicators while failing to provide a complete picture of their environmental impact.

Proposed Legal Framework

A comprehensive AI energy-disclosure regime should contain the following principles:

1. Mandatory Measurement

Large AI operators should measure energy consumption associated with significant AI activities.

2. Standardised Methodology

Regulators should establish common methodologies for calculating energy consumption.

3. Public Disclosure

Material aggregate information should be publicly accessible.

4. Regulatory Disclosure

Detailed technical information should be available to competent energy and AI regulators.

5. Independent Verification

Large operators should periodically verify their energy information.

6. Confidentiality Protection

Legitimate trade secrets and security-sensitive information should receive appropriate protection.

7. Enforcement

False or intentionally misleading disclosures should attract meaningful sanctions.

Conclusion

Energy disclosure obligations for AI operators represent an emerging area at the intersection of energy law, AI regulation, environmental law, corporate disclosure and consumer protection.

AI systems are not purely digital technologies. They depend upon electricity-intensive physical infrastructure. As AI deployment expands, governments increasingly require information concerning its energy consumption in order to plan electricity systems, evaluate environmental impacts, prevent greenwashing and protect affected communities.

The development of AI energy-disclosure law should therefore follow the principle that significant energy consumption should be measurable, transparent and accountable.

The most important legal lessons from cases such as Massachusetts v. EPA, Utility Air Regulatory Group v. EPA, Friends of the Earth v. Laidlaw, KlimaSeniorinnen v. Switzerland, Chamber of Commerce v. SEC, Liberty Energy v. SEC and Urgenda Foundation v. Netherlands are that environmental impacts can justify regulatory accountability, but disclosure obligations must also be based upon clear legal authority, rational methodology and procedurally sound rulemaking.

Ultimately, effective AI energy disclosure can create a regulatory chain:

Measurement → Disclosure → Verification → Transparency → Accountability → Energy Efficiency → Decarbonisation.

Therefore, energy disclosure should be regarded not merely as an administrative reporting requirement but as an important instrument of modern energy governance for the AI economy.