The Institute for Human Rights and Business (IHRB) published its briefing AI Infrastructure: Rights, Risks and Responsibilities (October 2026).
Human Level’s Take:
- AI is often discussed in terms of what it can do. But what about what it takes to make AI possible? As investment in data centres and supporting infrastructure accelerates, the human rights implications of this physical expansion are coming into sharper focus.
- The human rights footprint of AI starts well before any model is switched on. From mineral sourcing and construction to the land, water and energy data centres depend on, risks arise at every stage of the infrastructure lifecycle. Many risks are familiar from other sectors, but the speed of expansion raises questions about whether they are being adequately considered before decisions are made.
- That pace leads to a second question: what does AI infrastructure deliver, for whom, and at whose cost? Investment and jobs bring opportunities, but benefits and burdens are not always shared equally. Growing community opposition puts meaningful consultation, local benefits and whose voices shape development decisions front and centre, and for companies it is increasingly a project risk as well as a rights issue.
- Addressing this is complicated by who is involved. Technology companies, investors, developers, contractors and suppliers all play different roles and hold different degrees of influence. The expectation itself is already spelled out in the UN Guiding Principles and, for the largest companies, reflected in the CSDDD. The challenge is understanding where leverage sits and how it can be used to prevent and address impacts.
- If the risks are familiar, so are many of the tools. The way forward is less about reinventing the wheel and more about applying existing approaches early and effectively. Drawing on lessons from mining, energy and other infrastructure sectors, IHRB highlights lifecycle human rights impact assessments, meaningful community engagement, labour safeguards, responsible financing and clearer accountability.
- Some of these steps may already sit within existing corporate structures; others will need broader collaboration. With major investment and development decisions underway, IHRB describes a narrowing window, with the next 18 to 24 months shaping how AI scales for decades. That makes this a moment to embed human rights considerations from the outset, rather than address impacts once infrastructure is already in place.
Some key takeaways:
- AI may be digital, but its human rights footprint is increasingly physical: IHRB examines the rapid expansion of data centres and the land, water, energy, materials and labour underpinning them, citing estimates of nearly US$3 trillion in AI-related infrastructure investment planned through 2028. The scale of this expansion brings significant human rights implications into focus. A preliminary assessment presented at the first UN Global Dialogue on AI Governance in July 2026 projected that AI-related electricity demand could nearly triple by 2030, while water consumption could reach levels equivalent to the basic domestic needs of 1.3 billion people. These impacts span the full data centre lifecycle, from materials sourcing and construction to operation and decommissioning, and are unevenly distributed. While technology companies and investors benefit from expanding AI capacity, host communities may face rising energy costs, water stress, environmental degradation and disruption, often with limited long-term economic benefits or meaningful consultation. Responsibility is also spread across a complex ecosystem of investors, hyperscalers, landowners, contractors, equipment suppliers and energy providers. IHRB stresses that existing frameworks, including the UN Guiding Principles and OECD Guidelines, already establish expectations for companies to address human rights impacts across their operations and business relationships, while the amended CSDDD introduces binding requirements for the largest companies within scope. Yet existing EU data centre reporting requirements focus on resource efficiency, leaving issues such as land rights, labour conditions and community consultation largely unaddressed. The briefing therefore highlights the need to look beyond the environmental footprint of individual facilities to the wider human rights impacts and responsibilities associated with AI infrastructure expansion.
- The gap between promised benefits and realities on the ground is fuelling community opposition: IHRB highlights how the rapid growth of data centres is raising questions about who benefits from their development. A review of approximately 700 US data centres found that nearly half were located in areas with significant environmental burdens, many also characterised by high levels of social vulnerability. Employment benefits are another point of contention: an analysis of some US data centres found that more than 90% of jobs were created during construction, rather than long-term operation. Further upstream, demand for copper, nickel and other critical minerals connects AI infrastructure to global supply chains where poor labour conditions, including child labour, and environmental harm have been reported. Country examples illustrate how these risks manifest in different contexts. In the UAE, IHRB highlights risks of wage theft, passport confiscation and extreme heat for migrant construction workers, while in India, rapid data centre expansion has coincided with water stress and disputes over land acquisition. These concerns are increasingly translating into organised opposition. Between May 2024 and March 2025, more than 100 grassroots groups worked to block data centre projects worth an estimated US$64 billion across 28 US states, while projects have also faced delays or cancellations in countries including the Netherlands, Germany and Singapore. IHRB draws parallels with the oil, gas and mining sectors, but identifies the speed of AI infrastructure expansion as a distinguishing challenge: cumulative pressures on local resources and communities can emerge before companies and regulators have fully understood or responded to their consequences.
- The future of AI infrastructure will be shaped by how human rights are embedded in decisions about its financing and development: IHRB sets out sector-specific recommendations, noting that some can be implemented within existing corporate structures, while others require broader collaboration. For technology companies, these include publishing lifecycle human rights impact assessments, strengthening supplier standards and worker grievance mechanisms, ensuring meaningful community consultation, and negotiating binding Community Benefit Agreements covering utility costs, local jobs and revenue-sharing. The briefing also calls for board-level accountability and annual reporting on human rights and environmental performance. For investors and financial institutions, IHRB recommends embedding human rights due diligence into investment screening, financing conditions and monitoring, including requiring impact assessments for major projects and using investor leverage to strengthen accountability. Across the wider ecosystem, construction companies are encouraged to address labour risks throughout subcontracting chains and plan for responsible decommissioning; equipment suppliers to scrutinise critical mineral supply chains; and energy providers to protect household access to electricity. Drawing on its Dignity by Design framework and lessons from mining and energy, IHRB argues that companies need not start from scratch. Yet the window for action is narrowing: decisions on investment, regulation and corporate governance over the next 18 to 24 months could shape the human rights impacts of AI infrastructure for decades.
