Responsible & ethical AI use

We use AI to advance our mission responsibly, transparently, and with human oversight at every consequential decision. Every solution IOM builds or adopts is held to ten principles.

Humans decide; AI supports. No tool is the sole decision-maker on a person's case.

Ten principles

What responsible looks like, in practice

Do no harm

We watch for possible harm at every stage of using AI, from design to deployment. We assess risks early and provide clear channels for feedback and fixing problems.

Purpose & proportionality

We only use AI when there is a clear, justified need and it is the right tool for the job. Every use has a defined purpose and expected outcome.

Safety & security

We identify and manage safety and security risks throughout the AI lifecycle. Security controls and incident response are matched to the level of risk involved.

Fairness & non-discrimination

We use diverse, representative data and check for bias, so AI does not unfairly restrict people's rights.

Sustainability

We weigh the environmental, social and economic impact of AI. We choose energy efficient tools and responsible vendors, and assess impact where relevant.

Right to privacy

We protect people's privacy at every step: using data only for its stated purpose, lawfully, transparently, accurately and securely. We carry out privacy impact assessments (DPIAs) where required.

Human autonomy & oversight

People make the decisions, AI only supports them. Staff can check, control, override or shut down AI systems at any time. AI is never the sole decision maker on outcomes that affect people.

Transparency & explainability

We keep AI use clear and open at every stage. Decisions are documented, and people affected by them can get a meaningful explanation.

Responsibility & accountability

Every AI system has a named, accountable owner responsible for its governance, monitoring, fixing problems and reporting to the relevant oversight body.

Inclusion & participation

We involve diverse stakeholders and affected communities in designing and deploying AI. We build AI literacy and consult people meaningfully before launch.

In practice

Principles, lived in the field

An IOM colleague at a psychosocial support point — DO NO HARMSupport and safe referral pathways and stay close to the people we serve.

A staff member records data on a tablet in the field — RIGHT TO PRIVACYData is collected only when needed, and protected by design.

A migrant shares his story through an IOM app — TRANSPARENCYPeople understand, and consent to, how their information is used.

An IOM colleague engaging with a community — INCLUSION & PARTICIPATIONCommunities help shape and test what we build.

IOM colleagues in a responsible-AI training session — AI LITERACY5,000+ colleagues trained to build and use AI responsibly.

How it holds up

Backed by a lifecycle and clear ownership

The principles are enforced through a defined lifecycle, risk classification, and named accountable owners. They are aligned with leading international Responsible AI Use frameworks and practices.

A lifecycle, end to end

Discover, approve, design, deploy, monitor, then remediate or retire, with privacy by design and human-oversight points built in.

Risk-classified use cases

Every use case is rated Low to Very High, setting the assurance, monitoring and sign-off required, up to executive decision.

Named accountable owners

Each system has an owner answerable for its governance, monitoring and remediation, with oversight bodies reviewing the rest.

Every advance in capability, met with more oversight

From a phone in the field to agentic systems on the horizon, the same ten principles travel with everything IOM builds and adopts.

← Back to the overview