A newsletter on Responsible AI and Emerging Tech for Humanitarians
After six months of publishing Humanitarian AI Unpacked, we paused in April to listen. Through surveys, interviews, and polls, over 60 readers (thank you!) shared what they value most:
- Focus on depth of real-world case studies and practical applications,
- Feature learning opportunities for non-techies, tailored to humanitarian contexts, and
- Keep it simple - fewer visuals and more structured summaries
This new format is our response to you, and we hope it delivers.
This month, we are diving into one of your most requested topics: AI in Crisis Response. We will cover tools that are already helping humanitarians anticipate, plan, and respond to emergencies. From predictive models to MEAL chatbots, AI is no longer hypothetical, it is already operational.
But as humanitarian funding tightens, can these tools really help us do more with less? And do organisations have the time and resources to use AI safely, ethically, and well?
Let’s dive in.
Case Study: Mapping Vulnerable Populations with AI – International Committee of the Red Cross (ICRC)
When planning humanitarian aid, accurate population data is important but often missing. In conflict-affected or unstable areas, outdated census information forces humanitarian organisations to rely on crude estimates like counting rooftops from satellite images and multiplying them with the average household size. For the ICRC, this wasn’t good enough. As GIS analyst Thao Ton-That Whelan put it: A roof alone doesn’t tell you if it is a “administrative building, school, industrial building, or if it's a residential building”
To solve this, ICRC partnered with Swiss universities to develop POMELO, an AI model that generates fine-grained population maps down to 100x100 metres. Using open-source geospatial data (such as OpenStreetMap), building data and road proximity, and other indicators, the tool estimates population size and density, even in data-scarce environments. The results? In testing across sub-Saharan Africa, POMELO delivered more accurate predictions than traditional methods.
Yet turning the model into an operational tool wasn’t easy. While POMELO’s code was complex, the ICRC responded by building a Population Grid Hub, a user-friendly platform where staff could access, validate, and apply the population data. Field teams now use the hub to zoom into specific areas and draw boundaries to get tailored estimates which saves critical time during emergency planning.
ICRC made the POMELO code open-source to benefit the wider humanitarian community. But challenges remain: maintaining the model requires technical expertise the organisation is still building. Looking ahead, the team is exploring new methods like analysing night-time satellite images to detect population movement and infrastructure damage in near real-time.
The biggest lesson? Investing in AI can significantly improve crisis planning - but success depends on matching powerful tools with the human expertise to use them wisely. Can organisations collaborate to pool resources, avoid duplication and save costs?
Who’s Doing What
A snapshot of promising AI tools being used across the humanitarian sector.
A snapshot of promising AI tools being used across the humanitarian sector.
- Danish Refugee Council (DRC) – Foresight: Displacement forecasts
The Foresight model is a machine learning model that has been developed to predict forced displacement (IDPs, refugees and asylum seekers) at the national level 1-3 years into the future. - The Netherlands Red Cross – Automated Damage Assessment tool (ADA)
ADA uses deep learning to rapidly identify building damage from satellite images after disasters, cutting assessment time from weeks to hours and enabling faster response. - Catholic Relief Services – Measuring Resilience in Malawi (MIRA)
The MIRA project has devised a data collection and analysis scheme to measure and predict resilience among households prone to food insecurity one to two months out.
Editor’s Choice
Curated reads and resources our team found especially insightful this month.
📖 AI-Powered Analytics for Crisis Preparedness, Harvard T.H.Chan (2025) - The Crisis Sensitivity Simulator, an AI Powered model that uses historical data to model the multifaceted impact of crisis in specific countries, helping decision makers prioritise resources for crisis preparedness.
📺 Artificial Intelligence meets Humanitarians, NetHope webinar (2024) – Save the Children, Data Friendly Space and Amazon Web Services explore how AI enhances humanitarian efforts with real-life applications from Hurricane Beryl, the Sudan crisis, and the Lebanon crisis.
📖 SAFE AI Roundtable, FCDO (2025) - dive into the highlights of the SAFE AI project’s launch hosted by the FCDO with a total of 90 experts and leaders in AI and humanitarian action including AI’s transformational potential, private sector collaboration and community engagement.
Skill Up
Short, practical learning picks for practitioners - no tech background needed.
- UCL et al’s Designing and Deploying AI Tools to Support Humanitarian Practice A Practical Guide (free) - A guide designed for humanitarian professionals who want to integrate AI into their work responsibly and effectively.
- CDAC at al’s Community Crisis Intelligence (free) - Tools and practices that combine data from affected communities and responders with AI for more effective crisis mitigation, response or recovery.
- NetHope’s Humanitarian AI Code of Conduct (free) – A framework to guide the ethical and responsible use of artificial intelligence in humanitarian contexts.
- Coursera: “AI For Everyone” by Andrew Ng (free) – Beginner-friendly intro to AI concepts and social impact.
Podcast Spotlight
Voices from the sector on emerging tech deployment in humanitarian response.
Humanitarian AI Today, hosted by Brent Phillips with Alexandra Pittman, CEO of Impact Mapper, Suzy Madigan, Responsible AI Lead with Care International, Gary Forster, CEO of Publish What You Fund, and Linda Raftree and Quito Tsui from the Meryl Tech initiative.
The panel explores practical applications of AI in humanitarian crises, such as using machine learning for analysing qualitative impact data, automating monitoring and evaluation workflows, and supporting transparency through structured open aid data. The discussion features tools like ImpactMapper and highlights AI's role in synthesising massive amounts of unstructured data (e.g., PDFs and reports). It also critiques current AI adoption practices, urging caution around over-automation, ethical risks, and the importance of inclusive, context-aware implementation.
🕐 Run time: ~45 min
Upcoming Opportunities
Stay ahead of funding calls and events.
💰 UKHIH Humanitarian Rapid Research Initiative - Deadline: 4 June 2025 Call for proposals for 2025/26 Humanitarian Rapid Research Initiative (HRRI), offering up to £100,000 for an 8-month contract to deliver timely, actionable research that informs humanitarian responses to emerging crises.
💰 Humanitarian Innovation Programme (Innovation Norway) - Deadline: 13 June 2025 Call for proposals for funding (up to NOK 10 million per project) development and scale of innovative solutions that enhance humanitarian action.
🗓️ Third Sector: The Conference – When: 18-19 June 2025, London
While the event focusses on the third sector in the UK more broadly, it includes worthwhile sessions for humanitarians such as the state of tech, the digital code of practice and AI.
🗓️ AI for Good Global Summit – When: 8-11 July 2025, Geneva
Global event focused on ethical AI applications, including for humanitarian settings.
Nate Haken, previously the Fund for PeaceIt is much harder to predict the cascading impacts of the crisis after it occurs.
Disclaimer: The views expressed in the articles featured in this newsletter are solely those of the individual authors and do not reflect the official stance of the editorial team, any affiliated organisations or donors.