ADVANT Pulse
Your Labour & Employment News
Issue N4 - February 2025
Special Newsletter - AI in HR
As artificial intelligence (AI) continues to transform workplaces and is becoming increasingly integrated into employment processes such as hiring, employee monitoring, and employee evaluation. When using AI, companies already need to comply with regulation including data protection and labor laws. However, they will soon also need to ensure compliance with another regulatory framework the EU AI Act. The AI Act, published in August 2024, categorizes AI systems into risk levels, with high-risk systems subject to the most stringent requirements. With regard to these provisions, it will enter into force in August next year.
High-Risk AI systems under the EU AI Act In a employment context, the new regulation concerns foremost: a) AI systems intended to be used for the recruitment or selection of natural persons, in particular to place targeted job advertisements, to analyze and filter job applications, and to evaluate candidates. b) AI systems intended to be used to make decisions affecting terms of work-related relationships, the promotion or termination of work-related contractual relationships, to allocate tasks based on individual behavior or personal traits or characteristics or to monitor and evaluate the performance and behavior of persons in such relationships.
The most important requirements for high-risk AI systems at a glance The providers of high-risk AI systems bear the following obligations: Quality and risk management Technical documentation, record-keeping and logging obligations Consideration of accuracy, robustness, cybersecurity and accessibility during development Transparency and information obligations Registration in the relevant EU database and cooperation with the competent authority
Those who only deploy of high-risk AI systems generally have to fulfil fewer requirements than providers. However, there may be scenarios in which they can be subject to the same extensive obligations as the providers of high-risk AI systems.
Looking ahead, the AI Liability is poised to complement the EU AI Act. It aims to streamline legal pathways for individuals harmed by AI systems, including in employment related situations. However, the legislative process is still in its early stages and only rarely does a directive emerge from the legislative process in the form in which it was presented by the EU Commission.
Authors
Mickal d'Allende Partner ADVANT Altana [email protected]
Lakshana Cheetoo Associate ADVANT Altana [email protected]
Markus Knzel Partner ADVANT Beiten [email protected]
Gerald Mller-Machwirth Partner ADVANT Beiten [email protected]
Francesca Pittau Partner ADVANT Nctm [email protected]
Greta Quazzola Associate ADVANT Nctm [email protected]
1
AI: a look at France, Germany and Italy
1. Will the AI Act require any legislative initiative impacting national laws? As an EU regulation, the AI Act is (or will be) directly applicable in all member states without requiring formal transposition into national law. However, member states must establish national structures and authorities to monitor and enforce the regulation. France Following the adoption of the AI Act, the General Secretariat for European Affairs (SGAE) - the government agency in charge of relations between France and the European Union - is required to carry out an impact assessment of the AI Act in French Law. The purpose of this impact study is to identify the parts of the current legislation that need to be amended in order to comply with the provisions of the AI Act. The SGAE, in association with the various departments concerned by the AI Act will, then set a schedule in order to prepare the legislative measures needed to comply with the EU regulations. These texts will then be discussed by the Parliament. An administrative circular dated 22 March 2024 sets out the conditions for implementing European Union law.
Germany Germany is required to establish a national oversight structure for monitoring the AI Act by August 2, 2025, at the latest. The Federal Network Agency (Bundesnetzagentur) is expected to act as the central market surveillance authority, ensuring compliance with the regulation and conducting conformity assessments for high-risk AI systems. To implement the regulation effectively, Germany is working on a specific implementation law called the AI Market Surveillance Act (KIMarktberwachungsgesetz. This law aims to define clear responsibilities while minimizing bureaucratic hurdles to avoid stifling innovation. The AI Market Surveillance Act is expected to come into force in mid-2025, although it remains to be seen to what extent the Bundestag elections in February 2025 will have an impact on the further legislative process.
2
AI: a look at France, Germany and Italy
Italy The Government on April 23, 2024, approved a draft bill regulating the use AI in those sectors entrusted by the regulation to the Member States. The draft bill, outlines principles in line with those provided by the AI Act and requires the Government to adopt, within 12 months from the date the law comes into force, one or more legislative decrees to adapt national legislation in accordance with the AI Act. The draft bill is undergoing the parliamentary approval process.
2. How does AI impact Labor Law in each country? France French companies will, of course, not avoid this "revolution". This raises the question of whether the French labor code already provides them with the tools they need. AI today within companies Whatever the forthcoming transposition of the AI Act, if employers intend to use artificial intelligence to evaluate or control their employees, they must inform them in advance. The methods of control used have to be relevant to the objective pursued. Professional training cannot be ignored: employers have a legal obligation to ensure that employees are able to adapt to changes in their jobs and to maintain their ability to hold down their job. Jobs transformation Certain functions and roles may be replaced by AI. The question is to determine whether this is a simple change in working conditions or a change to the employment contract. Everything will depend on the scope of the changes and the pace at which they occur. Negotiations within companies with unions will be key. Furthermore, if the development of AI leads to the loss of certain jobs, companies may consider redundancies for economic reasons. If the company does not anticipate having to face economic difficulties due to the introduction of new technologies linked to AI - or having to reorganize in order to maintain its competitivity - the introduction of new technologies associated with AI should only be viewed as a technological change. French Labour Law is already operational Labor law therefore seems to be already effective in enabling companies to manage the emergence of AI. Let's hope that transposition of the text will not generate excessive constraints for companies in the EU.
3
AI: a look at France, Germany and Italy
Germany The integration of AI into the world of work is having a significant impact on German employment law. It brings with it a number of potential legal challenges: Data protection and automated decision making Article 22 GDPR: Automated individual decisions that have legal or similar ad-verse effects on employees are generally prohibited. For example, decisions on dismissals, promotions or bonuses may not be made solely by AI. Human control and final decisions remain mandatory. Risks of discrimination General Equal Treatment Act (AGG): AI systems must not make discriminatory decisions, for example in the application process or for promotions. Violations of the ban on discrimination can lead to claims for damages. Bias in algorithms: Employers must ensure that the AI systems they use do not produce unconscious biases that could lead to discrimination. Employment contract law Employees must always perform their work in person ( 613 BGB). The use of AI as an aid is permitted as long as essential tasks are not completely delegated to the AI. Clear regulations in the employment contract or in guidelines are advisable.
4
AI: a look at France, Germany and Italy
Co-determination rights of the works council Works Constitution Act (BetrVG): The use of AI systems, in particular for monitoring employee behavior or performance, is subject to co-determination by the works council (Section 87 (1) No. 6 BetrVG cf. below for further details). Employers must inform and involve the works council at an early stage ( 90 (1) No. 3 BetrVG). EU AI law and compliance High-risk AI systems, such as those often used in HR (e.g. for candidate selection or performance assessment), are subject to strict transparency, security and nondiscrimination requirements. Practical challenges Employers should implement clear policies on the use of AI in the workplace. These should include: o What systems can be used o The processes in which AI will be used. o How privacy and trade secrets will be protected. o A requirement for final human review.
Italy AI is already transforming the Human Resources sector with advanced tools for personnel selection, evaluation and management. For example, AI-based technologies and tools grant the possibility to optimize recruitment processes through automated CV screening systems (i.e. CV intelligence), talent scouting algorithms and predictive analysis of skills required in the job market. In addition, thanks to AI, companies can also identify the most suitable candidates and plan development and retention strategies, with an impact that continues to evolve. This is already a reality that poses new challenges for companies and employment lawyers too. In the meantime, as such subjects must manage the new regulatory landscape provided by the AI Act, it is also extremely important to understand how the AI Act interacts with the existing regulation. In particular, Italian law, as well as case law, has always prioritized employee protection, with strict regulations and limits concerning the surveillance and remote control over employees. These limits often require the employer to strike a balance between operational needs (such as ensuring productivity and security) and the rights of employees. Indeed, the main rules actually in force include: - Compliance with data protection in cases of automated decision-making process-es, according to the principles of GDPR;
5
AI: a look at France, Germany and Italy
- Information obligations if automated decision-making or monitoring systems are used by the employer, according to article no. 1 bis of Legislative Decree no. 152/1997 as recently amended; - Obligations of the principals towards workers of digital platforms (riders) accord-ing to article no. 47 bis of Legislative Decree 81/2015; - Regulation of remote controls over employees according to article 4 of Law no. 300/1970. In addition, certain employers have already implemented specific guidelines and/or codes of conduct to manage any interaction with AI-based technologies. Considering their general scope, even if such regulations also touch upon some aspects arising from the use of AI-based systems, the AI Act brings a layer of complexity for employers that requires a reshaping of the current framework.
6
Insight from France
The implications of AI under French Law: the employer's obligations towards its employees and staff representatives French labour law has not yet taken up the subject of artificial intelligence in order to regulate the labour relations that would be impacted by it. There is therefore no legislative framework that is strictly dedicated to Artificial Intelligence in French labour law. Nevertheless, the National Commission for Information Technology and Freedoms ("CNIL"), has issued publications on the subject, with the aim of preserving fundamental rights. Recommendations on the application of the GDPR to the development of AI systems have been published in order to help professionals to reconcile innovation and respect for the rights of individuals. The definition of a purpose for the AI system, a legal basis authorising the processing of personal data and a data retention period are thus mentioned as recommended practices. In any event, the Labour Code already sets out obligations applicable to employers who wish to introduce AI in their companies, and thus, at the heart of labour relations. As such, the importance of employees and their representatives is key. 1. The priority of respecting employees' individual rights and freedoms One of the fundamental principles of French labour law is that the employer is prohibited from imposing restrictions on the rights of individuals and on individual and collective freedoms that are not justified by the nature of the task to be performed or proportionate to the aim pursued. In this regard, employers should pay particular attention, when implementing any AI system, to respect the principle of non-discrimination (in particular regarding recruitment or performance evaluation tools), respect for the privacy of employees and the secrecy of correspondence (in particular for AI relating to employee surveillance and control methods). Consent is also at the heart of such implementations since there is an obligation in French labour law to inform employees in advance for any collection of information concerning them personally. In the event of an infringement of employees' individual rights and freedoms, aside from potential damages claims from the employees, the works council also benefits from a whistleblowing right.
7
Insight from France
2. The mandatory information and consultation of works council on the basis of several motives Although the agreement of the works council is not required, the introduction of AI in the company may have consequences that require it to be consulted. Indeed, the Labour Code requires that in companies with at least fifty employees, the works council is informed and consulted on issues concerning the organisation, management and general operation of the company, in particular, the introduction of new technologies and any major changes to health and safety conditions or working conditions. The works council may also be involved in other hypotheses, such as the implementation of AI resulting in a reduction of the workforce. AI can have consequences on staff management, through automated processing, involving mandatory information of the works council, or on the control of employees' activity as a technical means available to the employer, in which case the information and consultation of the works council are then required. Employers must consult the works council during the recurring consultations provided for by the Labour Code on the subject of AI as long as its implementation can influence the company's social policy, working conditions and employment, or the company's strategic orientations, which are all subjects of mandatory and recurring consultations between employers and staff representatives. These obligations are not to be dismissed as in the event of failure to consult or incorrect consultation (failure to meet deadlines, inaccurate information, etc.), the works council may request suspension of the measure planned by the employer, damages or even a criminal conviction for the offence of obstruction, leading notably to financial penalties. Thus, although the legislator has not yet taken up the subject of AI to guide employers through its implementation in the company, the fact remains that the provisions of the Labour Code; dating back a number of years, are already applicable and must be observed by employers. Moreover, it is also possible to negotiate collective agreements on this topic or to establish charters, like the IT charters that are already practiced by companies, in order to provide a framework for the practice and proper use of AI in companies.
8
Insight from Germany
AI and employee co-determination in Germany The rapid development of AI has far-reaching implications for the working environment. AI systems can perform not only repetitive tasks but also support complex decisionmaking processes such as the selection of job applicants or the implementation of complex social selection processes. As a reminder, AI systems that are used to analyze, filter, and evaluate applicants are classified as high-risk AI under the AI Act This also raises questions with regard to employee co-determination, as the use of AI can influence the organization of work and the working conditions of employees. Since the reform by the Works Council Modernization Act in 2021, explicit reference has been made to AI in several paragraphs in the Works Constitution Act (Betriebsverfassungsgesetz - BetrVG). However, the BetrVG does not contain a separate definition of what is meant by AI. The legislative reasoning, as documented, also contains no definition, meaning that disputes can be expected in future as to when employees' rights of co-determination will be relevant. The following AI-related constellations are explicitly regulated within the BetrVG, although they do not concern the subject matter of mandatory co-determination but are located in the area of information and participation rights. 1. Consultation of experts To provide works councils with the necessary expertise on the impact of AI in the working environment, Sec. 80 (3) sent. 2 BetrVG stipulates that an expert can be consulted on issues relating to the introduction and use of AI in the company. In this respect, the law presumes that an expert must be consulted in these cases. In principle, an inhouse expert (e.g. an employee of the IT department) should be consulted first and an external expert can only be commissioned if there is insufficient in-house expertise. In any case, the involvement of the expert requires an agreement with the employer, so that the works council may not act on its own authority. If such an agreement exists, the employer must bear the costs of such expert. 2. Information and consultation rights The employer must inform the works council about the planning of work processes and workflows, including the use of AI and submit the necessary documents to the works council, Sec. 90 (1) No. 3 BetrVG . This information must be provided at an early stage, i.e. at the planning stage and before the actual start of use of AI in the company. 3. Selection guidelines The consent of the works council is required in connection with selection guidelines (i.e. regulations on uniform specifications that affect personnel selection in the event of recruitment, transfer, regrouping or dismissal). Pursuant to Sec. 95 (2a) BetrVG, the works council's approval requirement also applies in the event that AI is used to draw up such guidelines.
9
Insight from Germany
4. Right of co-determination In addition, the right of co-determination pursuant to Sec. 87 (1) nos. 1 and 6 BetrVG are particularly relevant. These genuine co-determination rights are characterized by the fact that the works council can demand the conclusion of a works agreement and thus employers may not unilaterally introduce and operate the corresponding measures (here: the introduction of AI systems). These are subjects of mandatory codetermination. Measures taken by the employer without the proper involvement of the works council are legally ineffective. The introduction of AI in the company in violation of co-determination rights can thus be prohibited by the works council - if necessary, by way of an interlocutory injunction. 4.1. Right of co-determination pursuant to Sec. 87 (1) No. 1 BetrVG (regulatory behaviour) Employers who draw up regulations on the order and conduct of employees in the workplace must involve the works council in accordance with Sec. 87 (1) No. 1 BetrVG and conclude a works agreement on this. By contrast, measures relating to work behavior, i.e. regulations that determine how work is to be performed in the exercise of the right to issue instructions, are not subject to co-determination. Corresponding specifications of the work obligation do not lead to a co-determination right of the employer (cf. BAG, decision of 23 August 2018 docket number 2 AZR 235/18). Accordingly, the instruction to use or not use AI to perform the work duties would not be subject to codetermination (cf. ArbG Hamburg, decision of 16 January 2024 docket number 24 BVGa 1/24). 4.2. Right of co-determination pursuant to Sec. 87 (1) No. 6 BetrVG (technical monitoring equipment) According to the case law of the Federal Labor Court, this co-determination requirement is to be understood in a broad sense, so that not only the introduction and use of technical equipment intended to monitor the behavior and performance of employees leads to the right of co-determination, but also such technical equipment that is suitable for such monitoring (cf. BAG, decision of 10 December 2013 docket number 1 ABR 43/12). AI systems are designed for the input of data and its processing, whereas it can be assumed that this (user) data will also be stored. The data stored accordingly is then also suitable for monitoring employees (e.g. when did who provide which input?). However, the employer must also be able to access the data stored by the AI. And there is no such possibility of access if the AI system is not set up on the employer's servers, but the employees access the AI via their own (private) access points (cf. ArbG Hamburg, decision of 16 January 2024 docket number 24 BVGa 1/24). This overview shows that the planning for the introduction of AI systems already entails information rights for the works council and that the works council can have a say in the introduction and use of AI systems. Case law will have the task of further defining the relationship between AI and co-determination.
10
Insight from Italy
1. The use of AI in Human Resources: opportunities and regulatory Limits The AI Act has introduced specific restrictions for the Human Resources sector, identifying certain practices as prohibited and classifying many applications as "high risk" under a risk-based approach, imposing strict compliance measures. Among the prohibited practices, one of the main restrictions concerns the use of AI for emotion recognition in the workplace based on biometric data. Indeed, article 5, 1, letter f) of the AI Act expressly bans emotion inference in the workplace unless there are specific medical or security reasons. Additionally, the regulation prohibits social scoring (article 5, 1, letter c), meaning the evaluation of employees based on personal data or information unrelated to the work-place, such as, for example, those provided by social networks. These bans aim to prevent the risk of business decisions, such as hiring or promotions, being influenced by factors unrelated to professional features. Indeed, the improper use of these technologies could violate the employee's rights to privacy and dignity, relying on subjective and potentially discriminatory metrics. It goes without saying that, even without automated tools, the "social media persona" is already evaluated and awareness on the topic is a must for job seekers. In the context of said of high-risk systems, the AI Act (article 14) expressly requires human oversight in the decision-making processes to ensure fairness and non-discrimination. This should address the risk, highlighted in the regulation, that automated systems replicate underlying biases, disfavoring some categories and reinforcing existing limits in the diversity of the organizations.
11
Insight from Italy
In such context, the legislator has indeed embedded a reference to ethical and responsible use of AI technologies, providing guidelines for the development of "humancentered" systems, with measures that can be mostly voluntarily applied by deployers (i.e., employers), for example through the implementation of codes of conduct. This aspect, along with the importance of transparency towards employees, is our focus below. 2. The importance of transparency and codes of conduct in the AI Act The AI Act requires employers using AI technologies and tools to ensure the explainability and traceability of automated decisions so that candidates and employees can understand the underlying logic. In this context, transparency serves as a safeguard for employees' rights. According to article 50 of the AI Act, employers have to provide clear and accessible information to employees regarding the use of AI, including in decision-making processes. Additionally, the regulation includes specific measures to ensure that automated decisions are understandable and challengeable, along with audit and monitoring mechanisms to prevent discriminatory effects. In particular, employers will need to declare the use of AI, conduct risk assessments, and put in place safeguards to protect employees' rights. In this context, a key element of the regulation is the role of codes of conduct. The AI Act encourages companies and organizations to develop voluntary codes of conduct that establish principles and procedures for an ethical use of AI, especially in the HR sector. Although not binding, such codes (article 95) serve as essential tools to ensure respect for human dignity, non-discrimination, and data protection. They can be adopted by AI system deployers or organizations representing them, involving stakehold-ers, civil society, and academia. Codes of conduct aim to overcome possible pitfalls in AI-based systems through measures such as: Promoting AI literacy for developers and users; Designing inclusive AI systems with diverse development teams and stakeholder
involvement; Evaluating and preventing the negative impact of AI on vulnerable individuals,
including gender equality and accessibility for people with disabilities.
12
Insight from Italy
3. Protection against discrimination and pay transparency a perspective on the AI Act As above, since algorithms are often trained on biased data sets, one of the main risks associated with the use of AI technologies and tools in HR is that they may potentially perpetuate or amplify systemic discrimination, even unintentionally, in performance evaluations, hiring decisions, or disciplinary actions, disproportionately affecting certain groups of employees. Indeed, employee selection and performance evaluation models could reflect biases already present in historical data, leading to discrimination based on gender, ethnicity, age, or disability. To mitigate this risk, the AI Act also obliges the employers to ensure the quality and representativeness of the data used to train algorithms (article 26, 4). Additionally, article 14 states that all automated decisions in the workplace must be subject to human oversight to prevent opaque and uncontestable decision-making processes. An additional safeguard for the employee will come from the pay transparency framework, as regulated by the EU Pay Transparency Directive (2023/970). This Directive (which has an implementation deadline of June 2026) will compel employers to provide detailed information on salary criteria to prevent unjustified pay disparities between men and women. The principle of pay transparency is particularly relevant in a context where algorithms could influence salary dynamics unfairly. A key aspect of the Directive is the reversal of the burden of proof (article 18): in cases of non-compliance with pay transparency obligations, it is the employer's responsibility to prove that no discrimination has occurred. This principle could also impact on the use of AI when applied to salary metrics and reviews, requiring companies to demonstrate that their automated systems do not produce discriminatory effects. Will we be able to balance technological innovation with the need to ensure a fair and transparent labor market by adopting the preventive measures defined by European legislation?
Follow us!
ABOUT US
ADVANT is a distinctly European law firm association, uniquely positioned to help you navigate the complex and nuanced legal and commercial landscape of Europe www.advantlaw.com
CONNECT
Stay tuned with the most recent ADVANT news and content by following us on LinkedIn and Instagram
ADVANT* Altana | Beiten | Nctm.
*ADVANT is an association of legally and financially independent law firms, including Altana Socit d'Avocats in France, Beiten Burkhardt Rechtsanwaltsgesellschaft mbH in Germany and Nctm Studio Legale in Italy. All rights reserved 2025.
13
Imprint This publication is issued by: ADVANT Altana, a French socit d'exercice libral responsabilit Iimite, with registered address at 45, rue de Tocqueville, 75017 Paris, France; ADVANT Beiten Ganghoferstrasse 33, 80339 Munich, Germany Registered under HR B 155350 at the Regional Court Munich / VAT Reg. No: DE11218811 For more information see: www.advant-beiten.com/en/imprint; and ADVANT Nctm, an ltalian Associazione Professionale, with registered address in Via Agnello 12, 20121 Milano, Italy.
ADVANT Altana, ADVANT Beiten and ADVANT Nctm are members of ADVANT, an association of independent law firms. Each Member Firm is a separate and legally distinct entity, and is liable only for its own acts or omissions. This publication cannot replace consultation with a trained legal professional.
Your Employment Advantage
www.advantlaw.com
14
