English - Español - Català - Français - Nederlands
Weather

–°

Hollywood, FL

Mandatum HeraldVox Populi, Vox Dei
Barça…
Peñarol…
Inter Miami…


AI in Humanitarian Work: Useful Tools, Strict Limits

by hbmadero | Aug 24, 2026 | Technology | 0 comments

Nov 2024ICRC guidelines
4Core ideas
3Named risks
20Expert interviews

Artificial intelligence is arriving in humanitarian work, in logistics, in medicine and in the analysis of data from crises. The organizations that serve people in the worst moments of their lives are adopting it carefully, and they are putting their rules in writing.

The Red Cross sets its guidelines

In late November 2024 the International Committee of the Red Cross (ICRC) issued its own AI guidelines. Its delegate for technological diplomacy, Philippe Stoll, said the organization cannot afford a fast, experimental approach, because it serves highly vulnerable people. The ICRC defines AI broadly, as any system that performs tasks normally requiring human thinking, planning or reasoning, so the rules cover much more than chatbots.

The guidelines rest on four ideas:

Core principles

AI must respect humanity, impartiality, neutrality and independence.

Real need

It is adopted only where there is a real need, not to look for problems an AI tool might solve.

Data protection

Sensitive personal data must be protected.

Local adaptation

Tools must be adaptable to local conditions, since many models are trained mainly on Western data.

What it is already used for

The ICRC points to Meditron, an open language model led by the Swiss Federal Institute of Technology in Lausanne (EPFL) and trained on carefully selected data to support clinical decisions. It also cites supply-chain software, developed with EPFL, that helps manage shortages and surpluses of equipment, a problem that directly affects patients.

The risks are concrete

Data breaches

In 2022 a cyberattack exposed the personal data of hundreds of thousands of people receiving aid.

Irrelevant advice

A medical tool that suggests treatments unavailable locally “makes no sense.”

Fragile tools

Tools that are not robust enough could harm the very people they are meant to help.

Four principles from the research community

A report hosted by UN OCHA’s ReliefWeb, based on scholarly sources and 20 expert interviews, proposes four principles for using AI with vulnerable people: weigh benefits against risks and avoid AI where possible; use systems grounded in the local context; include and empower local communities; and audit algorithms.

The test that matters: across these documents the message is the same. In humanitarian work a tool is judged by whether it serves the person in front of it, protects their data and respects their context. Speed and novelty come second.

Sources: Geneva Solutions, “Red Cross adopts AI guidelines to contain risks” (December 4, 2024); ReliefWeb, “Artificial intelligence principles for vulnerable populations in humanitarian contexts.” Image: Earth at night from the International Space Station, NASA, public domain, via Wikimedia Commons.

Keep reading