Today, 4 June 2026, we are in London for the conference “AI and the Future of Social Welfare: Recent Developments and Future Prospects”, hosted at the London School of Economics.
The two-day conference brings together researchers, practitioners, and policy experts to discuss one of the most urgent questions facing welfare systems today: how can artificial intelligence be used to strengthen social protection without weakening trust, fairness, and human dignity?
The event is jointly sponsored by the International Network for Social Policy Teaching and Research, the London School of Economics, the University of California Berkeley Center for Comparative Welfare State Research, and the University of Maryland School of Public Policy. Across the programme, participants are examining how AI is already changing social welfare, from digital service delivery and automated administration to predictive analytics and new forms of risk assessment.
For social protection, the promise is clear. AI can help make services faster, improve access to benefits, identify emerging risks, reduce administrative burdens, and support public servants in focusing their time where human judgement and care matter most. But the risks are just as real. If poorly governed, AI can reproduce bias, obscure decision-making, create new forms of exclusion, and make welfare systems feel more distant from the people they are meant to serve.
We are represented at the conference by Dr Chris De Neubourg, Director of SPRI Global, and Dr Stiene Praet, Senior Researcher. They are presenting our paper “Bridging the Gap: AI Governance for Equitable Adaptive Social Protection in the Face of Climate Shocks”, co-authored by Stiene Praet, Míriam Carrera Manzano, Julia Karpati, and Chris De Neubourg.
Our paper focuses on the use of AI in adaptive social protection, particularly in contexts affected by climate shocks. Governments and development partners are increasingly exploring AI for early warning systems, vulnerability assessments, targeting, and emergency scale-up. These tools can help anticipate needs, identify affected populations, and deliver support more quickly when droughts, floods, heatwaves, or other climate-related shocks place pressure on households and public systems.
Yet this is precisely where governance matters most. Climate-sensitive social protection often operates in urgent, high-stakes, and resource-constrained settings. Decisions may need to be made quickly, based on imperfect data, while the people affected may already face poverty, displacement, limited connectivity, or weak access to public services. In such contexts, AI is not simply a technical tool. It becomes part of the decision-making architecture that determines who receives support, when, and on what terms.
Our paper maps international AI governance frameworks against the practical risks that arise in real-world social protection systems. It then proposes a governance toolkit for policymakers, donors, and practitioners, helping them assess when AI should be used, what safeguards are needed, and how to ensure that AI-supported responses remain effective, accountable, and equitable.
For us, the central question is not whether AI belongs in social welfare. It already does. The real question is whether we can govern it well enough to ensure that it serves people, especially those most exposed to climate risks. At a time when both digital systems and climate shocks are expanding rapidly, this discussion could not be more timely.
We look forward to the exchanges at LSE and to continuing our work on AI governance, adaptive social protection, and climate resilience.