Building a responsible humanitarian approach: the ICRC's policy on artificial intelligence

Summary

The ICRC sets out how it will treat artificial intelligence inside its own work. The policy is written from the organization's humanitarian mandate and Fundamental Principles, not from a product plan or a compliance checklist. It was developed with humanitarian expertise, existing international AI standards, and outside expert comment, and it is aligned with the ICRC's 2024 to 2027 Institutional Strategy. It covers staff education and humanitarian applications. The ICRC is cited here as a source and as an example of careful policy writing. That citation does not mean the ICRC endorses, partners with, or is affiliated with AI Concerns.

The principles it rests on

In our words.

1. Mandate before tool: the work stays humanitarian, even when the method is new.
2. A human-centred approach: people, not systems, remain the point of the work.
3. Safe exploration: staff may learn and try, within the bounds of the mandate.
4. Consultation: the text was formed with operational expertise and external review.
5. Named limits: the authors say what the document cannot settle.

Scope

AI use across ICRC operations, including staff education and humanitarian applications, anchored in the organization's mandate and Fundamental Principles.

Why it is exemplary

It is written from principles and mandate rather than from product. It keeps people at the centre. It states its own limits in public, and it was developed in consultation rather than in isolation.

Limits the authors state themselves

“Given the constantly evolving nature of AI, this document cannot possibly address all the questions and challenges that will arise in the future.”

ICRC, 2024. The stated aim is to help staff "safely explore its humanitarian potential" while remaining "anchored in a purely humanitarian approach driven by our mandate and Fundamental Principles."

Concerns in our register it speaks to

• Right to human decision-makerD19
• Migrant and refugee rightsD19
• Civilian-harm amplificationD11
• Debate: ban, moratorium, meaningful human controlD11

Frameworks it relates to

• International humanitarian law (IHL)
• ICRC Fundamental Principles (as named by the authors)
• ICRC Institutional Strategy 2024 to 2027 (as named by the authors)

How a reader might use it

• A school

  Read it as a model of writing from purpose and limits before a classroom tool is chosen.

• A health system

  Read it as a model of a human-centred policy that says, in public, what it cannot yet settle.

• A public body

  Read it as a model of consulting expertise and of stating uncertainty without delaying a first, bounded policy.

Source

ICRC publication page

International Committee of the Red Cross, Building a responsible humanitarian approach: the ICRC's policy on artificial intelligence, 18 November 2024.