← All research

Who is responsible for safe AI deployment at work?

Providers, management, business teams, IT, worker representatives and prevention specialists each see part of the system. Safety depends on explicit coordination.

Responsibility cannot be delegated to the model

The provider knows the model; the employer organises work; users see discrepancies in activity; representatives and prevention professionals observe collective and health effects. No actor holds all the information alone.

Responsibility becomes fragile when everyone assumes someone else checked safety, or when the final decision is attributed to a technical recommendation.

Organise decision loops

Safe deployment names the owner of each risk, defines stop criteria and makes incidents visible. It distinguishes who proposes, authorises, uses and can suspend.

Governance is not a one-off meeting. It is a loop connecting design, decision, use, reporting and correction.

03

Why it matters

Distributed responsibility without coordination can easily become an absence of operational accountability.

04

Source

IA Santé Travail — cadre de gouvernance

Open the source