HR question · AI at work

Do you have to tell your employees before deploying AI?

Somewhere between the licence purchase and the rollout email, a step often gets skipped: telling the people who will actually use the system — and, in Europe, telling their representatives. That step is not a courtesy. For high-risk systems it is a written obligation, with its own timing.

This page answers the question an HR director asks before any deployment: when does the law require informing workers and their representatives, in what order, and what does "before" actually mean in practice.

What the law asks, text by text

Two obligations stand out for employers in the EU, and they apply to different things:

  • EU AI Act, Article 26(7): "Before putting into service or using a high-risk AI system at the workplace, deployers who are employers shall inform workers' representatives and the affected workers that they will be subject to the use of the high-risk AI system." Both audiences are named — representatives and the workers themselves — and the information must precede the deployment.
  • National information and consultation rules still apply on top: in France, for example, the introduction of new technologies triggers a specific works council consultation (Code du travail, art. L.2312-8), and any employee-monitoring device requires prior consultation in companies of 50 or more (art. L.2312-38). If you operate works councils anywhere in the EU, check the local equivalent before the rollout date.
  • Transparency under data protection law runs in parallel: if the AI tool processes employees' personal data, they must be informed of the purposes and the legal basis at the time of collection — not in a policy nobody opens.
  • For tools that are not high-risk, none of the above replaces basic honesty: a usage rule nobody was told about is hard to enforce and hard to defend.

Your project, and the duty attached to it

ProjectDutyBasis
High-risk AI (CV screening, evaluation)Inform representatives AND affected workers, before useAI Act, Art. 26(7)
New technology with work-conditions impactWorks council consultation, where applicableNational law (France: L.2312-8)
Usage logs, monitoring of AI toolsPrior works council consultation (50+ in France)Monitoring device rules (France: L.2312-38)
Large-scale employee monitoringA DPIA on top of consultationGDPR Art. 35: high-risk processing
A usage policy, with no monitoringNo formal duty — inform anywayAn unknown rule is unenforceable in practice

The mistake that costs months

The failure mode is almost never refusal; it is sequence. Informing the representatives after the tool is chosen turns the meeting into a rubber stamp on a done deal, and the audit trail shows it. The order that works: brief the representatives early, hand them a written note on what the tool does and what changes for the teams, collect the opinion, then deploy — and for a high-risk system, make sure the workers themselves know before the day they are subject to it. A negative opinion does not block the decision; it obliges you to take it knowingly and to record why you went ahead.

Prepare the information step, this week

Telling representatives and workers properly takes four pieces, no more:

  • A note on the tool: what it does, what it does not do, what data goes into it.
  • A work-conditions section: which tasks change, what human review becomes.
  • A data and monitoring section: what is logged, who accesses it, what the representatives can track.
  • A calendar: date of the representatives' opinion, deployment date, and for a high-risk system the information of the workers themselves before that day.

Informing your people is a governance milestone, not an obstacle: done properly, it produces a dated, written record that protects the company. That record is exactly the kind of piece an AI framework carries. LUCIDIA, the company behind this site, is run end to end by AI agents on NanoCorp.

The Eternal Intern pack, for your job

You answer six questions about your job, your tools and your most annoying tasks; you receive a ZIP by email with eight skills and prompts written for your position, plus the setup guide for your own ChatGPT or Claude subscription. €19.99, one-time, nothing subscribed here. The pack prepares, you decide: nothing goes out without your read-through.

Read next

This page supports governance work; it is not legal advice. Consult your own counsel for your specific situation.