HR in the Know 12/2024

AI systems used by employers will be subject to disclosure and consultation

The Use of AI

In recent years, the development and a significant increase in the use of algorithms and artificial intelligence technologies is quite noticeable across various sectors. AI helps employees save the time by supporting i.a. activities such as drafting messages, creating summaries, searching for information, translating, and even other creative functions like designing presentations and generating ideas. More and more often decisions related to recruitment, so-called derecruitment (employee dismissal), performance evaluations, work optimisation, access to training, determination of salary, or employee promotions are taken with the use of algorithms or AI systems. This was supposed to bring objectivity and process optimisation, but it also poses significant risks, especially when the outcomes of AI systems are not monitored by decision-makers. In such circumstances, the lack of transparency can lead to actions which in utmost cases, might result in unequal treatment of employees, particularly through discriminatory algorithms or unlawful collection and use of sensitive data.

AI Act

The European Union has undertaken numerous legislative measures to protect employees from the adverse effects of artificial intelligence. On 12 July 2024, the so-called AI Act – formally known as Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (artificial intelligence act) and amending other EU legislative acts („AI Act”) — was published. The priority of the designed regulation is to define a regulatory framework for development, enforcement and use of AI within the territory of Member States. The other key task is making sure that AI solutions used in the European market are safe, transparent, traceable, non-discriminatory, environmentally sustainable and supervised by humans. The new regulations state specific obligations on both the providers and users of algorithms, including new information requirements. The information obligation under the new rules includes, in particular, information to employee representatives and individual employees concerned that a high-risk AI system will be used in relation to them (taking into account their purpose and the type of decisions and actions made by algorithm). Thus, the informational solutions adopted by the AI Act are extensive encompassing not only trade unions but also employee representatives. Of course, the obligations for system providers and users will include more than just informational duties including i.a. the requirement to monitor and in appropriate circumstances respond to the activities of these systems, develop detailed procedures for risk classification and estimate risk related to AI use, continuously monitor these risks, and establish a risk management system.

Formally, AI Act has been adopted as a regulation, hence its provisions will be directly applicable within the territory of the EU. The Regulation entered into force on 1 August 2024, though, due to its complicated character the moment of application has been postponed in time until 2 August 2026 (subject to some exceptions).

Amendment to the Trade Union Act

In parallel with the ongoing EU's legislative efforts, certain independent legislative activities have also been undertaken in Poland aligning with overall EU legislative trends. In another attempt to amend the Trade Union Act in relation to AI, a committee draft was submitted to the Polish Parliament (Sejm) on 29 April 2024 (i.e. before the adoption of the AI Act). It aims at expanding employers' obligations to inform trade unions about issues related to the use of algorithms or artificial intelligence systems by employers. The draft was relatively quickly (already on 11 June 2024) referred to the first reading in committees – to the Committee on Digitalization, Innovation and Modern Technologies, the Committee on Economy and Development, and the Committee on Social Policy and Family. However, at this stage the work had stopped, likely due to i.a. the fact that work on the EU AI Act was being finalised at the same time.

According to the argumentation for the draft amendment, the current Polish legislation, and consequently the Trade Union Act, is not adapted to the rapidly changing technological realms and reality. The need to expand the catalogue of information results from the increasing use of algorithms or AI systems. Currently, Article 28(1) of the Trade Union Act imposes certain informational obligations on employers only insofar as necessary for trade union activities, particularly regarding four essential pieces of information: working conditions and remuneration rules; employer’s activity and economic situation related to employment and anticipated changes; condition, structure and anticipated changes in employment and efforts to maintain the headcount; and actions that may significantly affect the organisation of work or employment foundations. The amendment proposes to extend this information obligation to include the fifth point, stating that employers will be required to provide information regarding: 5) the parameters, principles, and instructions on which algorithms or AI systems that influence decision-making are based and that may impact working and pay conditions, access to employment, and its retention, including profiling. The draft Act is to enter into force 14 days after its publication.

Effects of the Proposed Legislative Changes

According to the authors of the amendment, the expanded informational rights granted to workplace trade unions are intended to facilitate their oversight of artificial intelligence and allow for increased transparency in employer actions related to human resource management, as well as to eliminate potential errors resulting from the use of algorithms. Given that the draft amendment does not impose consultation obligations towards trade unions, the authors emphasise that it will not affect productivity-enhancing activities, and therefore, it will not negatively impact business competitiveness. One of the issues highlighted as significant in the proposal, is remote work and the use of IT systems to monitor the ongoing results of this work associated therewith. The authors emphasise that employees have the right to understand the principles of using technological tools employed in the workplace, and providing this information helps avoid deliberate circumvention of labour code provisions. Unfortunately, the project in its current form may raise many practical concerns, as it does not define the algorithms or artificial intelligence systems to which the informational procedure is to apply. Potentially, any such systems could affect the conditions of work or pay of employees (or access to employment for candidates), but the practical question is where to draw the line. Moreover, it is worth mentioning that the unionisation rate in Poland, which according to CBOS (Polish Public Opinion Research Centre) data from 2021, stands at approximately 10.5% of employed persons. This means that unfortunately, the informational obligation imposed only on unionised employers does not constitute a comprehensive regulation that increases the protection, or at least awareness, of the majority of employees (persons performing paid work) in Poland. Trade unions mostly operate at larger employers or those with historical links to the public sector, leaving a huge number of small private sector employers, where it is more difficult to conduct an audit, out of the reach of such regulations.

With technological advancement, including AI systems, their role in supporting work increases, which in turn raises the need to protect employees from their negative effects. We can certainly expect many changes in labour law which will be result of extensive discussions on AI systems held at the European Union forum. Considering the legislative activity in this field thus far, it is expected that work on this project will take relatively long. The first step will probably be changes in the Trade Union Act, which may, in a later version, take into greater account provisions contained in the AI Act published on 12 July 2024. For now, the scope of informational and consultation duties in the AI Act and the proposed draft amendment to the Trade Union Act remain diversified, which could cause further practical issues. The AI Act provides for an obligation to consultation with employee representatives and employees regarding high-risk AI systems used in the workplace, while the amendment to the Trade Union Act includes only informational procedures and only with regard to trade unions. Adapting the Polish Act's wording with the EU regulation would undoubtedly help prevent the emergence of parallel, hard-to-reconcile obligations imposed on employers.