Local, Everywhere: The blueprint for a Humanitarian AI transformation

Published on December 14 2025

The slogan ‘Local, everywhere’ of the 2026 Global Plan of the International Federation of Red Cross and Red Crescent Societies (IFRC) captures the essential blueprint for the bottom-up development of Humanitarian AI: technology must be grounded in community knowledge to be effective everywhere it is needed. Humanitarian AI is not about shiny new tools for their own sake. It is about using technology to capture, preserve and apply the Red Cross and Red Crescent Movement’s greatest asset: the lived experience and expertise of millions of volunteers and staff across the globe. With more than 17 million volunteers and over half a million […]

The slogan ‘Local, everywhere’ of the 2026 Global Plan of the International Federation of Red Cross and Red Crescent Societies (IFRC) captures the essential blueprint for the bottom-up development of Humanitarian AI: technology must be grounded in community knowledge to be effective everywhere it is needed.

Humanitarian AI is not about shiny new tools for their own sake. It is about using technology to capture, preserve and apply the Red Cross and Red Crescent Movement’s greatest asset: the lived experience and expertise of millions of volunteers and staff across the globe.

With more than 17 million volunteers and over half a million staff, the Movement generates an extraordinary amount of insight about crises, communities, local risks, and practical solutions that work on the ground. AI can help turn this dispersed and tacit knowledge into a shared resource: a common good for humanitarian action.

The good news is that bottom-up development of Humanitarian AI is ethically desirable, technically feasible, and financially affordable.

Why bottom-up AI is ethically desirable

Knowledge shapes identity. Preserving knowledge is closely tied to maintaining dignity and, ultimately, our shared humanity. For thousands of years, communities have passed on knowledge through stories around the fire, local legends and oral histories, and later through schools and universities. Today, AI accelerates how knowledge can be captured and organised, something anyone can feel when interacting with ChatGPT, DeepSeek, and other AI platforms.

The future we should strive for is one where knowledge is developed and preserved in its context, reflecting culture and societal specificities. For humanitarian action, local knowledge is a matter of practical necessity. Dealing with a crisis – whether caused by a tsunami, drought, conflict, or epidemic – unfolds in a specific local context.

Locally embedded AI can also help address some of the ethical concerns that surround AI today:

Bias and mis/disinformation: When AI systems are anchored in curated humanitarian knowledge, with clear attribution of sources, it is easier to identify and correct harmful biases and errors. Instead of having often abstract discussions about the ethics of major AI models, we can address the ethics of individuals and communities, which are reflected in local AI systems.
Trust: AI developed and governed by humanitarian organisations, rather than opaque commercial platforms, can be aligned with humanitarian ethics and principles from the outset. Trust, as an inter-human condition, will be embedded in individuals and institutions behind AI models, not in technical platforms. We trust (or not) each other, not machines. When it comes to major AI platforms, we can (or not) trust the owners of these companies, such as Elon Musk or Sam Altman. Through practical AI platforms, the humanitarian community can help regain trust worldwide as a critical aspect for the future of humanity.
Responsible reuse: Humanitarian AI models, weights, and datasets can also be utilised to align commercial AI systems with core humanitarian use, including non-harmful inferences. OpenAI, Claude, and DeepSeek can retrain their models in order to ensure legal and ethical compliance with humanitarian principles. Providing such a service could also serve as a source of income for humanitarian actions worldwide.

Why is bottom-up AI technically feasible?

Humanitarian AI is technically feasible:

Open-source language models: Powerful models can now be deployed on relatively modest hardware, especially when trained for specific domains, such as humanitarian response.
Rapid creation of AI agents: It is already possible to set up simple AI agents in minutes. The real work lies in training and refining them with high-quality, real-world knowledge over time.
Edge and mobile capabilities: Increasingly, AI can run directly on mobile devices or low-cost servers, reducing dependency on large AI factories or expensive cloud infrastructure.

Humanitarian AI can benefit from the current shift from a focus on computing power (e.g., the number of NVIDIA cards) to training AI models based on the knowledge and data of citizens, communities, and universities worldwide.

In addition, most humanitarian use cases, such as decision support, scenario planning, knowledge search, language assistance, translation, or guidance for volunteers, do not require massive supercomputing resources. Instead, they require high-quality, well-structured data and knowledge, as well as robust mobile-first tools that can run in low-resource environments, in the hands of both volunteers and staff.

Done this way, AI will not only help preserve knowledge but can also increase the security and resilience of digital solutions used by the Movement.

Why are AI solutions financially affordable?

AI is quickly becoming a commodity technology. Open-source AI tools can be deployed at relatively low cost. The Movement, including the IFRC, can make AI even more affordable by, among other things, developing a Humanitarian AI toolkit consisting of open-source servers trained on humanitarian data, humanitarian weights, knowledge graph models, and AI agent applications.

In addition, the Movement can create sustainable funding models by ensuring that whenever commercial chatbots and platforms utilise Humanitarian AI, the generated income should belong to the individuals and institutions that contributed knowledge to the development of these models. Such income could help fund operations, local branches, or further knowledge work, turning AI from a cost centre into a source of financial support.

While considered a global public good to be utilised by the Movement, commercial uses of Humanitarian AI should always benefit the communities and institutions on which knowledge AI solutions are developed.

How AI can support IFRC’s Renewal priorities

Bottom-up AI, anchored in humanitarian knowledge, can directly support the IFRC’s Renewal priorities:

Deepening localisation and accountability: AI systems tailored to cultural, societal and policy contexts can support locally led action and decision-making. Transparent attribution of sources and reasoning can strengthen accountability, both within the Movement and towards the communities it serves.

Sharpening humanitarian focus, influence, and impact: AI can help synthesise evidence-based insights from field reports, assessments and local data. It can amplify the voices of local actors, ensuring that their perspectives are visible in national, regional and global debates.

Intensifying collaboration and trust: Shared local AI knowledge repositories can become a focal point for collaboration between the IFRC Secretariat, National Societies, and local branches. Co-creating these repositories with communities can deepen trust and engagement, and strengthen volunteer networks.

Accelerating digital transformation: AI can support scenario planning, operational analysis, and real-time decision-making based on data from volunteers and branches. It can help surface ‘hidden gems’ of tacit knowledge within the Movement, practical know-how that is often undocumented but crucial in times of crisis.

Becoming a more innovative, agile, and lean secretariat: AI can help the Secretariat better organise and use its own knowledge across projects, operations, and crisis responses. AI capacities will need to extend beyond technical teams to include officials working on programmes, policy, and emergency response, so that AI supports their daily decisions rather than sitting on the margins.

Anchoring AI in the humanitarian fundamental principles

A bottom-up approach to AI can directly reinforce the Fundamental Principles of the Red Cross and Red Crescent Movement.

Humanity: AI that builds on individual and community knowledge respects the dignity and lived experience of people affected by crises. It focuses on supporting human free will, not replacing it.

Impartiality: By drawing on a wide range of sources and clearly attributing them, humanitarian AI can help represent diverse perspectives while maintaining the Movement’s impartiality.

Neutrality: Models can be designed and monitored to avoid taking positions on controversial political, racial, religious, or ideological issues, reflecting the Movement’s commitment to neutrality.

Independence: If critical decisions and daily operations rely on proprietary, opaque AI platforms, the Movement’s independence could be endangered. Thus, developing local AI solutions grounded in humanitarian principles and local knowledge could be crucial for preserving the Movement’s independence.

Voluntary service: Recognising the importance of volunteers’ knowledge in developing Humanitarian AI can reinforce the ownership of ideas and actions among volunteers. Acknowledgements such as “key AI knowledge contributor” could reinforce motivation, recognition and pride in service.

Unity: AI solutions and knowledge bases should be accessible to all relevant actors within a country across branches and levels of the National Society.

Universality: Humanitarian AI can help build a shared pool of global humanitarian knowledge, treated as a public good. Local insights, once attributed and consented to, can inform better responses worldwide.

Where to start: Small steps and agile deployment

Realising the potential of a non-hyped, humanitarian-centred AI approach does not require a massive, one-off project. It can begin with small, focused pilots on codifying and preserving knowledge from the IFRC Secretariat and National Societies, and then extending it to local branches. From there, successful pilots can be scaled step by step to other parts of the Movement and wider humanitarian community.

The key to success will not be the computing power of AI models, but anchoring AI solutions in local cultural, social and political realities that reflect the ethics and experience of the people using them.

If the Movement can combine its long humanitarian tradition with thoughtful use of AI, it can not only respond more effectively to crises today, but also preserve the knowledge that future generations will need to save lives tomorrow.