.png)
Imagine a community affected by severe flooding. Families need shelter, food, and medical support. Humanitarian workers receive reports from different locations, in several languages, and sometimes with contradictions. Humanitarians have to quickly identify urgent needs and decide where help is most needed.
Artificial intelligence (AI) can help humanitarians make sense of this information. But using AI in crisis contexts is not the same as using it to draft an email or summarize meeting notes. So what makes AI ‘humanitarian ’, and how is it different from the commercial tools many of us already use?
There are many different definitions of humanitarian AI. For us at Data Friendly Space, we believe that it refers to the use of artificial intelligence to support humanitarian action, from disaster preparedness to emergency response and recovery. It can help humanitarian organizations analyse information, translate messages into different languages, identify patterns, and anticipate potential risks.
What makes it humanitarian is not only where it was used but how it is used. Humanitarian action is guided by four principles: humanity, neutrality, impartiality, and independence. AI used in this work should respect them too. A tool that ranks needs mainly by the volume of reports, for example, may overlook communities that are harder to reach or less connected. It could also help a response team review reports from flood-affected communities. It could group similar needs and highlight areas that require further attention. That would undermine impartiality.
Humanitarian AI also has to work under conditions that commercial products rarely face: poor connectivity, local and minority languages, fast-changing situations, and information that is incomplete or contested.
AI cannot replace human judgement. Its outputs can be inaccurate, incomplete, or biased. Humanitarians must review and question the results before acting on them.
Commercial AI tools such as ChatGPT or Claude are mainly developed for business, productivity, and everyday use. They can help people write content, analyse data, or automate tasks. Humanitarian AI focuses on applying these capabilities to humanitarian challenges.
The difference is partly about purpose and partly about design. A humanitarian-focused tool is shaped by choices about which sources it draws on, how transparent it is about where its answers come from, and how it handles sensitive data.
Commercial AI tools can support many tasks, including writing, research, and data analysis, and humanitarian organizations may use them in their work. However, they must consider data privacy, accuracy, and whether a tool is suitable for the situation.
Let’s go back to the example of a flood. A general-purpose chatbot asked, “Where are the most urgent?” may give a fluent answer. But it may draw on unverified or outdated web content or even invent details, also known as a hallucination. It may not show where the information even comes from unless prompted by the user. A humanitarian-focused tool should draw on verified humanitarian sources and cite each claim. The GANNET Virtual Assistant offers an example of AI designed specifically for humanitarian work. Its Virtual Assistant helps humanitarian organizations analyse reports and find information from verified humanitarian sources. It can process documents in multiple languages and provide answers with citations. These citations help users trace information back to its original sources. Like any AI tool, GANNET has limits. Its answers are only as complete as the sources available, and coverage may be thinner for some languages or regions. That is why human review remains a core part of how it is used.
Data is a particular concern in humanitarian situations. Information about affected people, such as their location, ethnicity, or legal status, can put them at risk if mishandled. Guidance such as the IASC Operational Guidance on Data Responsibility in Humanitarian Action sets out how to manage these risks, and it applies whichever AI tool is used.
The right AI tool depends on the problem an organisation needs to solve. A general-purpose tool may be enough for some tasks. Other situations may require tools designed primarily for humanitarian data and crisis analysis.
Before adopting any AI tool, ask:
Humanitarian AI is useful only when it serves that goal: it should be transparent about its sources, careful with people's data, and always checked by human judgement. As humanitarian organisations explore new AI tools, the priority should be clear: finding ways to use technology that support better decisions and respond to people's needs.
Want to see how this works in practice? Explore GANNET or read one of our crisis reports. Follow DFS for the next article in this series.
