AI meets ESG: the need for environmental awareness in adopting AI

Businesses considering adopting Artificial Intelligence (AI) are urged not forget about its negative environmental impact. The rapid adoption of AI presents significant opportunities, but it also introduces new challenges, particularly concerning its environmental footprint. As businesses seek to align their operations with Environmental, Social, and Governance (ESG) objectives, it is crucial to address the environmental impact of AI and find ways to mitigate its negative effects.

AI & ESG

AI has become a cornerstone of digital transformation, contributing to industries ranging from healthcare, retail, transportation and logistics, education, energy and utilities to finance. The potential benefits are clear: AI can streamline operations, improve efficiency, and even drive sustainability in sectors like energy management. Its environmental cost, however, is often overlooked.

AI systems rely on vast amounts of energy, raw materials, and data. The lifecycle of AI, from development to deployment, creates significant environmental challenges. These include the energy-intensive process of training AI models, the hardware required (e.g. semiconductors and GPUs), and the operation of AI systems through data centres. Research suggests that by 2027, AI could account for billions of cubic meters of water usage and could consume as much energy as an entire country.

The EU AI Act: What businesses need to know

The EU AI Act is the most comprehensive legislative effort to regulate AI, and while it does address some environmental issues, its primary focus is on safeguarding fundamental rights. The Act acknowledges the positive potential of AI for environmental sustainability but does not explicitly address the environmental costs AI might incur. Instead, it places a stronger emphasis on how AI systems impact health, safety, and human rights.

Nevertheless, several provisions within the Act begin to touch upon environmental protection, providing a framework for future action and encouraging businesses to start thinking about sustainability in AI development.

The key provisions of the EU AI Act relevant to environmental protection include:

• Fundamental rights impact assessment (Article 27): For AI systems deemed “high risk,” providers are required to conduct risk assessments, including those related to fundamental rights. The scope of these rights, however, remains unclear. While environmental well-being is recognised as a fundamental right under EU law, it is not clear whether it is explicitly included within the definitional scope of the EU AI Act. This ambiguity presents both challenges and opportunities for businesses, particularly when determining whether the environmental impact of their AI systems needs to be factored into these assessments, and this consideration alone would require extensive knowledge of fundamental rights law.
• Harmonised standards and standardisation deliverables (Article 40): Article 40 outlines the development of harmonised standards for high-risk AI systems or general-purpose AI models. The Commission will issue standardisation requests, which will include requirements for improving the resource performance of AI systems, such as reducing energy consumption. For businesses, this provision highlights the importance of staying informed about new standards and aligning their AI systems with these expectations to ensure compliance and promote long-term sustainability.
• Exemption from conformity assessment (Article 46): Article 46 provides for exemptions from the usual approval process for high-risk AI systems under specific circumstances, including for environmental protection. In urgent cases (e.g. for public safety or environmental emergencies), AI systems can be used without the prior completion of a conformity assessment, provided they undergo post-approval assessments. Businesses should be aware of this provision, as it could offer a degree of flexibility in deploying AI systems that serve the environment. Businesses, however, must be prepared for the process of post-deployment approval and ensure they are following the correct protocols to avoid non-compliance.
• Regulatory sandboxes (Article 59): The EU AI Act also establishes regulatory sandboxes (Article 59), which allow AI systems to be tested in a controlled environment using personal data. These sandboxes can be particularly useful for AI systems developed for public interest purposes, such as environmental protection and energy sustainability. For businesses involved in developing AI solutions with environmental applications, participating in a regulatory sandbox could provide valuable insights into system performance, risk mitigation, and regulatory compliance.
• Reporting serious incidents (Article 73): Under Article 73, providers of  high-risk AI systems must report serious incidents to authorities, including serious environmental harm. AI providers must report incidents within 15 days with stricter deadlines for more severe incidents. For clients, this provision emphasises the need for robust incident response plans. Businesses should ensure that their AI systems are equipped to monitor environmental impacts and establish reporting mechanisms to comply with the Act.
• Voluntary codes of conduct (Article 95): Article 95 of the Act promotes the creation of voluntary codes of conduct for AI systems. These codes aim to establish standards that encourage AI providers to minimise environmental impact, promote inclusivity, and ensure that AI technologies are developed responsibly. Businesses should be proactive in adopting and developing these codes, as they can help ensure AI systems are aligned with environmental and social goals. Additionally, adopting these voluntary codes may provide a competitive edge by showcasing commitment to sustainability.

Navigating the ESG challenge: practical solutions for businesses

For in-house counsel and businesses focused on sustainability, balancing AI adoption with environmental responsibility is crucial. Here are some practical steps businesses can take to ensure AI development aligns with the “E” in ESG:

• Assess the full environmental impact: When adopting AI, look beyond the benefits it can offer to ESG goals. Consider energy consumption, water usage, and other environmental impact associated with operating AI systems. This may include indirect effects, such as the carbon emissions or resource depletion caused by hardware manufacturing.
• Consider sustainable infrastructure: Consider assessing the sustainability credentials of the infrastructure supporting AI systems. Businesses may wish to consider service providers that demonstrate a commitment to environmental responsibility, such as the use of renewable energy sources and the implementation of water-efficient technologies. Additionally, thoughtful distribution of AI workloads can help to optimise resource use and minimise environmental impact.
• Implement sustainable AI design: Ensure that AI systems are developed with sustainability in mind. This can include designing systems to be energy efficient, optimising algorithms to reduce energy consumption, and prioritising the use of renewable resources in hardware production.
• Monitor and track environmental impact: Consider establishing processes to measure and track the environmental footprint of your AI operations. This will not only help in managing the impact of existing systems but also prepare for potential future regulations, including those related to sustainability reporting.
• Due diligence and supply chain responsibility: Businesses should consider incorporating ESG considerations into their supplier contracts, requiring suppliers to adopt sustainable practices in their AI development and operations. By fostering a sustainable supply chain, businesses can mitigate risks and improve their overall ESG performance.

Moving forward: the path to responsible AI adoption

AI can drive significant advancements in business operations and ESG goals, but it is essential to consider both the positive and negative impact of AI on the environment. By taking a proactive approach to sustainability, businesses can align AI adoption with their ESG objectives and mitigate its environmental risks. Whether through improved design, responsible infrastructure choices, or supply chain diligence, businesses can ensure that AI contributes positively to both their operations and the planet.

The article was co-authored by Helena Siebenrock.

For more information on AI and ESG, contact your CMS client partner or these CMS experts.