The European Commission publishes its White Paper on AI

The European Commission published a White Paper on Artificial Intelligence, setting out policy options for a future EU regulatory framework rather than proposing binding law itself. Alongside it the Commission opened a public consultation, inviting member states, industry, civil society and the public to respond before it decided how to legislate; the paper explicitly excluded military uses of AI from its scope.

The document’s central proposal was a risk-based approach, under which obligations would fall only on AI applications judged “high-risk” by two cumulative tests: whether the AI was deployed in a sector where significant risk could be expected — the paper named healthcare, transport, energy and parts of the public sector — and whether the specific use within that sector was itself likely to create significant risk to safety, consumer rights or fundamental rights. Remote biometric identification, such as facial recognition in public places, and AI used in recruitment were flagged as always high-risk regardless of sector. For applications meeting that bar, the White Paper sketched mandatory requirements later formalised in the EU AI Act: documented and representative training data, record-keeping sufficient to trace a system’s decisions, disclosure of a system’s capabilities and limits to users, provable robustness and accuracy, and human oversight ranging from pre-approval of outputs to a real-time stop function.

Building an ecosystem of trust is a policy objective in itself, and should give citizens the confidence to take up AI applications and give companies and public organisations the legal certainty to innovate using AI.

— European Commission, White Paper on Artificial Intelligence

The White Paper committed to no deadline and no legal text; its risk tiers, consultation model and vocabulary nonetheless carried through largely intact into the Commission’s AI Act proposal the following year, making it the first clearly identifiable sketch of what became the EU’s binding, risk-tiered regime for AI.