Building Bridges 2026: Financing AI and Its Implications for People and Resources
Updated: 3 hours ago
Building Bridges 2026 brought more than 2,500 participants from 78 countries to Geneva from 6 to 8 October. Under the theme “Investable Solutions to Global Challenges”, the seventh edition focused largely on climate, energy, infrastructure and nature. Yet artificial intelligence, described in the programme as “the double-edged driver of the transition”, surfaced in session after session.
In the opening discussion, Patrick Odier, Chair of Building Bridges, raised the question of why AI is attracting so much capital while climate and environmental protection still struggle to be adequately funded in some areas. Behind that question sits another one: what does the AI buildout mean for the people who work for it and live alongside it?
A value chain finance cannot yet see
The most direct answer came from a session built on the work of the United Nations Environment Programme Finance Initiative (UNEP FI) on AI. Titled “Banking perspectives on social and human rights impacts of artificial intelligence”, it set out from an uncomfortable premise:
“Banks are lending into the AI value chain faster than they are building the tools to assess it.”
The physical infrastructure of AI, including data centres, carries significant human rights risks, yet, as the session framing put it, “while the financial sector increasingly recognizes environmental impacts, social considerations lag behind.” For banks, the question is therefore practical: how to engage their data centre and AI clients on human rights, what to ask, what to look out for and where they have leverage.
To answer it, the session followed the value chain itself. Four tables, each hosted by a human rights, labour or technology expert, moved from the minerals in servers to banks’ own use of AI. Among the speakers were Isabel Ebert, Co-Lead of the B-Tech Project at UN Human Rights; Ioana Tuta, Senior Adviser at the Danish Institute for Human Rights; Sher Verick, Coordinator of the International Labour Organization’s Programme on Digitalization and AI; and Joana Pedro and Alice Anders of UNEP FI.
One of the most human links in that chain is labour. At the table on work, participants asked whether AI will displace jobs, which parts of the workforce are most vulnerable, and whether these effects fall differently on women and men.
Less visible still are the workers inside AI supply chains, often described as “hidden workers”. Little is known about their working conditions, and they are less likely to be organised in trade unions, although this is changing in some countries. That invisibility shaped the questions raised at the table: who the intermediaries are, what conditions workers face, and which complaint mechanisms and remedies they can turn to at national and global level.
Data centres: where AI meets communities
Data centres, by contrast, are where AI becomes physical, and where its demands on shared resources become concrete. At the high-level plenary “AI and the Net Zero Race”, Alix Lebec, Founder and CEO of LEBEC, cited a United Nations University study estimating that by 2030 the water footprint of data centres’ electricity use could equal the basic water needs of 1.3 billion people.
Those resources are often local, and so are the tensions. At the UNEP FI table on data centres, participants raised concerns that in some jurisdictions these facilities are treated as priority projects and are not subject to the same regulations as other developments. Rather than leaving each lender to deal with this alone, one suggestion was for banks to engage regulators collectively and push for a level playing field.
The discussion then moved from risk to what communities should gain. Participants argued that communities should take part in decisions on where and how data centres are built, for instance through multi-stakeholder structures. They called for real benefit sharing, designed for the community rather than presented to it as a selling point, and asked how communities hosting a data centre might also benefit from AI itself. They also stressed the need for grievance mechanisms that go beyond environmental issues, for example when water studies prove wrong, and that capture social harms such as gender-based violence during construction.
What a tangible benefit might look like was illustrated at the plenary by Devrim Celal, Chief Flexibility and Marketing Officer at Kraken. He described a scheme in which residents near wind turbines received electricity discounts whenever the turbines were turning. When the offer was opened across Britain, some 40,000 communities expressed interest within days. Data centres, he suggested, need a similar logic.
Social licence as a financial variable
For investors, this is increasingly a question of risk as well as fairness. Nili Gilbert, Chair of the Investment Committee at the David Rockefeller Fund, described social licence as one of the scarcest resources in the AI buildout, and urged investors to engage communities as early as possible and to plan for their benefits. The cost of failing to do so is already measurable: according to Gilbert, over $200 billion of data centre-related projects have been delayed in the United States this year alone.
Part of the difficulty, Gilbert noted, is that many institutional investors do not yet know how exposed they are to AI. That exposure runs through public equities, infrastructure, private credit and venture capital, and reaches portfolios indirectly through rising power and commodity costs and through effects on the workforce. The levers differ accordingly, from shareholder and bondholder engagement to direct questions on water use, siting, power needs and community engagement in infrastructure investments.
The UNEP FI discussion arrived at the same point from a different direction. There, community opposition delaying data centre projects was cited as the clearest business case for taking social impacts seriously, alongside reputational risk and regulatory risk, the latter so far concentrated mainly in Europe.
Participants at the UNEP FI session noted that corporate disclosure on how AI is developed and deployed is still at an early stage, and that some impacts, such as long-term effects on mental health or on the information ecosystem, may only become clear years from now. Financial institutions are, in effect, making judgement calls under uncertainty.
Uncertainty does not mean there is nothing to ask, and participants identified questions banks can already put to their clients. At board level, which policies and processes govern the risks AI poses to people, which committees oversee them, and how they are monitored and enforced. On transparency, how the company discloses its use of AI, including its human rights due diligence and the role of human oversight. Across the organisation, whether staff are trained on these risks, and how leadership communicates its human rights commitments.
Our perspective
At Inclusive Intelligence International, we work to integrate international human rights standards, including the UN Guiding Principles on Business and Human Rights, into the governance and financing of AI. Read through those standards, the discussions in Geneva point to three priorities.
The first is to identify and address adverse impacts across the whole value chain. Human rights risks do not begin or end with an AI model; they run from mineral extraction to data centres and the use of AI by financial institutions themselves. Under the Guiding Principles, responsibility also extends to impacts linked to a company’s business relationships, which makes financial institutions part of the chain of responsibility rather than only observers of their clients’ risks. The second is meaningful participation: workers and communities affected by AI and its infrastructure should be able to influence decisions, not simply be informed once they are made. For workers, this includes attention to the effects of automation on employment and the need for reskilling. The third is access to remedy, through grievance mechanisms that work in practice, not only on paper. In our view, the benefits of AI should reach society as a whole, not only a narrow group of actors.
To conclude, we turn to a question raised by Ronald Weidner, Advisor at X, the Moonshot Factory: “If we just want to optimize the current systems, it becomes an accelerant to extraction. What we really have to do is train AI. What do we want it to value?”
Comments