AI Act And New Guidelines Become Operational: Tools For Trade Union Agency?

The EU AI Act’s new transparency rules are now becoming operational. On 20 July 2026, the European Commission published its final Guidelines on the implementation of Article 50, ahead of the transparency obligations becoming applicable on 2 August 2026.

For European trade unions, the importance of these Guidelines goes beyond Article 50 itself. Their potential becomes clearer when Article 50 is read together with Article 26(7), which contains a specific workplace information obligation, and with existing rights to information, consultation, data protection and collective bargaining.

Three layers of workplace transparency

The AI Act can be understood as providing three complementary layers of workplace transparency.
The first is visibility. Article 50 makes certain forms of AI use visible to individuals. People must, for example, be informed when they are directly interacting with certain AI systems. Individuals must also be informed when they are exposed to emotion-recognition or biometric-categorisation systems. Article 50 also contains transparency requirements for certain AI-generated or manipulated content.

The second is collective prior information. Article 26(7) addresses a different situation. Before an employer deploys a high-risk AI system in the workplace, it must inform workers’ representatives and the workers concerned that they will be subject to its use.

The third is actionable transparency. Existing information and consultation rights, the GDPR, national labour law and collective bargaining can potentially allow workers’ representatives to go beyond simply knowing that AI is present. They can ask how it is being used, what it does, how it affects workers, who is responsible for it and what safeguards are in place.

This distinction is important. Information by itself does not correct the power imbalance that can arise when employers introduce algorithmic management. Transparency becomes much more useful when trade unions can connect information rights with consultation, negotiation and other forms of worker’ participation.

From notification to meaningful information for trade unions

Article 50 should not be interpreted as a general right for trade unions to discover every algorithm used by an employer. Its scope is narrower.

The Commission identifies four main situations covered by Article 50: interaction with AI; AI-generated or manipulated content; emotion recognition or biometric categorisation; and certain deepfakes and AI-generated public-interest content.

This means that an AI system operating silently in the background could be very important for workers without triggering an Article 50 notification. A system could, for example, rank workers, allocate tasks or assess performance without employees directly interacting with it.

But the difference between Articles 50 and 26 also creates an opportunity.

Article 26(7) applies before a high-risk workplace AI system is deployed. Workers’ representatives and affected workers must be informed in advance.

Article 50 however operates at a different point. It makes certain interactions with, or exposure to, AI visible to the individual worker.

Consider for example, an employer introducing an AI-enabled HR application. Employees begin receiving Article 50 notices when using it, but their trade union or works council received no prior information about the system. This does not automatically mean that the employer has breached Article 26. But it gives workers’ representatives a concrete reason to ask whether the system is high-risk and, if it is, whether the employer complied with its Article 26(7) obligations.

In this sense, Article 50 can act as a practical check on Article 26, also because Article 26(7) has an important limitation. It requires employers to inform workers and their representatives that they will be subject to a high-risk AI system. However, it does not itself establish a comprehensive right to an explanation of the system. The new Article 50 Guidelines cannot change Article 26 or create additional legal rights. However, their practical approach to transparency supports a broader principle: Information about AI should be sufficiently meaningful to allow the person receiving it to understand the situation and respond appropriately.

This approach receives useful support from the latest OECD work on artificial intelligence in finance[1]. A shop steward gains relatively little from being told that a bank, insurance company or other employer “uses AI”. What matters is what the system actually does and how it affects work, whether by allocating tasks, monitoring performance, detecting fraud, generating content or influencing staffing decisions.

The Article 50 Guidelines are interesting in this respect because the Commission explicitly intends them to turn a relatively abstract transparency obligation into practical expectations. They do not perform that task for Article 26(7), but they demonstrate the kind of operationalisation that may ultimately also be needed for workplace transparency.

Why this matters particularly in Nordic finance

These issues are especially relevant to trade unions representing workers in financial institutions. AI use is already widespread in finance. The OECD describes applications ranging from vendor-provided tools and open-source systems to automated trading, risk-management systems, natural-language processing and large language models. This matters because AI does not always arrive in the workplace as an obvious new “AI system”. It can be built into ordinary software, supplied by an external vendor, added to an existing HR or customer-management system, or integrated into processes employees have used for years.

This is particularly important in Nordic countries. According to Eurostat[2], Nordic countries are well above the EU average in organisational AI use, with Nordic companies among the EU leaders in AI adoption. Denmark recorded the largest increase in 2025, followed by Finland. This means that EU rules on workplace AI are likely to have particularly significant practical consequences for Nordic workers.

The experience of the Finnish financial sector illustrates why. Recent Finnish research[3] shows that employees in banking and insurance already expect AI to change the skills they need. Most employees anticipate significant reskilling and continuous learning as AI transforms work and competence requirements.

At the same time, attitudes are not simply negative. Many workers see opportunities for AI to improve productivity, particularly in insurance. This optimism, however, exists in parallel to considerable concern. Some employees report that AI has increased their workload, while many remain uncertain about its longer-term consequences. Expectations of future job displacement are especially notable: more than four in ten financial-sector employees expect displacement, compared with just about three in ten among Finnish trade union Pro’s white-collar workers overall.

This Nordic example demonstrates why workplace AI policy should not be reduced to a debate between being “for” or “against” technology. Workers may see considerable benefits in AI while simultaneously wanting influence over how it changes jobs, workloads, skills and employment prospects.

Article 50 can make certain uses of AI visible. Article 26(7) can make the deployment of high-risk workplace AI a matter of collective prior information. Existing labour law, data-protection rights and collective bargaining can then help turn that information into influence over how AI is used at work. This is how transparency may become a tool of collective bargaining.

Anna-Delia Papenberg, Head of EU Affairs at NFU

[1] OECD - AI In Finance
[2] Eurostat, 20% of EU enterprises use AI technologies
[3] Finnish Institute for Occupational Health, Gen AI in the financial sector survey