Informing Workers Before High Risk AI Is Used at Work
Article 26 includes an explicit requirement that deployers inform workers before high-risk AI systems are used to monitor or evaluate their work. This is a transparency obligation, but it is also rooted in labour rights and dignity. Workers have a right to know they are being monitored or assessed by an AI system. This requirement applies to all employing organisations, including public authorities, that deploy high-risk AI systems in an employment context.
The informing requirement is not met by a single notice buried in an employee handbook or posted on an intranet that almost nobody reads. The regulation requires that workers are informed in a clear, understandable manner and have the opportunity to ask questions or raise concerns. Many organisations interpret this to mean a face-to-face briefing session, a clear written notice, or a combination of both.
Consider a logistics company deploying an AI system to monitor driver performance, flag speeding events or unsafe following distances, and feed performance ratings into promotion and compensation decisions. The company must tell drivers, clearly and in advance, that this system is in use, what data it collects, what decisions it influences and how drivers can contest decisions they believe are unfair. Telling drivers only after the system has been collecting data for six months is a breach of Article 26.
Content of the notification
The notification to workers should cover several points. The deployer must identify the AI system and its purpose: “We are implementing an AI-powered performance monitoring system that will be used to evaluate your work”. The deployer must explain what the system does in non-technical language: “The system will track time spent on tasks, review emails for certain keywords and flag unusual patterns”. The deployer must be clear about how the system influences employment decisions: “Performance ratings from this system will be considered in annual performance reviews and may influence salary decisions”. The deployer must describe workers’ rights: “You have the right to ask for human review of any decision the system influences” or “You may request an explanation of any performance rating the system assigns”.
The notification should be delivered in a way that gives workers genuine notice. Email can work for office workers, but shop floor workers or field staff may need in-person briefings. Workers should be given time to read and digest the information and should be told how to raise questions or concerns. A union representative or worker representative should be involved where applicable.
Objection and contest rights
Workers should be told that they can contest decisions influenced by the AI system. This might mean requesting human review of a system recommendation or asking for reconsideration of a performance rating. Organisations should establish a clear process for contesting AI-influenced decisions and for appealing a contest decision to a manager or committee. Documentation of this process should be shared with workers.
Timing and updates
The notification must occur before the system is deployed. If the system’s purpose or capability changes materially, workers must be notified again. For example, if an AI system initially used only to categorise emails is expanded to influence promotion decisions, workers must be informed of this change.
Notification best practices
Organisations should communicate with workers in multiple ways. A single email notification is insufficient because many employees may miss or misunderstand it. Consider a combination of approaches: an email clearly marking the subject line “Important: AI Monitoring System Implementation”, a written guide in simple language explaining the system, a live information session where workers can ask questions, and posted materials in common areas. For unionised workforces, union representatives should be involved early and should help deliver the message.
The notification should be timed carefully. Telling workers about AI monitoring just as they are leaving for the weekend or during busy periods reduces engagement. Schedule notifications during working hours and allow time for workers to absorb the information and raise questions.
High-risk versus other systems
This requirement applies specifically to high-risk systems. Organisations deploying lower-risk systems do not have the same explicit notification requirement, though transparency and good employment practice may suggest notifying workers anyway. Knowing which systems fall into high-risk categories is therefore an important first step.
Organisations should document that they made a deliberate decision about which systems are high-risk and which are not. This documentation supports compliance if regulators inspect AI governance. The decision should be based on the system’s categorisation under the EU AI Act and should be reviewed periodically as the regulation’s interpretation evolves.