Skip to content
PKF News

You are here:

AI in the workplace - The risks call for guidelines and operational instructions

The chatbot ChatGPT was launched by OpenAI, a US American company, in November 2022. Since then, such so-called large language models (generative artificial intelligence, AI for short) have played a role in a very wide variety of situations in everyday life - a broad range of diverse conversations have been held with them and miscellaneous tasks are carried out by them. The question that has now arisen is: under what general conditions may employees likewise use AI applications in their day-to-day work?

Legal barriers to the use of AI 

AI users can have texts or other contents - such as tables, graphics or also program codes - created in accordance with their specifications by text-based dialogue systems. These many different possible applications as well as the rapid and low-threshold use to solve, in some cases, complex tasks makes the application of AI for professional purposes likewise attractive. Consequently, however, problems may arise in relation to employment rights, data protection legislation and copyright and we have considered these in the section below.

Employment rights

From an employment law perspective, first of all, it is questionable whether the application of AI is compatible with an employee’s personal performance obligation under Section 613 of the German Civil Code. According to this, employees are obliged to provide their services in person. In this respect, the question that then arises here is whether the use of AI to carry out a task rather constitutes the use of a technical aid (such as, e.g., a word processing program) or an application that operates independently. In the latter case, this would then be comparable to transferring the activity to another person. On the basis of the current state of the art and the German interpretation of the law it is thus possible to answer that AI has to be regarded as merely a work instrument and auxiliary tool. This correlates with the fact that AI has not been assigned a separate legal personality and a final check by the user remains absolutely necessary. Therefore, according to the current understanding, the obligation to provide a service in person is thus maintained. 

Employers are able to order or ban the use of AI, within the scope of their reasonable discretion, under their right to give instructions in accordance with Section 106 of the Industrial Code (Gewerbeordnung, GewO). Should an employer decide to ban or to restrict the use of AI applications then, in the event of the breach of such an obligation by an employee, the employer would have the standard measures under employment law at its disposal.

Furthermore, the employer should think about how and to what extent an employee should disclose that their output was produced with the help of AI. The use of AI-produced output could moreover have far-reaching consequences for the employer in the context of copyright or data protection. That is why employers should clearly define the use of AI-produced output for all parties. 

It is also anticipated that AI applications will be used in human resources departments. Particular care should be taken to ensure that no so-called profiling is carried out. 

Please note: A decision that is based solely on AI is not possible. AI may make recommendations, however, the final decision has to be made by a human.  

Data protection legislation

From the perspective of data protection legislation, the relevant question is: to what extent may personal data be used in AI applications? If an employee enters personal data as defined in Art. 4(1) GDPR into text-based dialogue systems then the problem related to data protection legislation can be solved via the consent of the data subject in the processing procedure in accordance with Art. 5(1)(a) GDPR. To this end, however, knowledge of the precise nature of the data processing and the specific purpose of the processing will be necessary - this is something that mostly cannot be determined in advance when it comes to AI tools. 

Recommendation: In this respect, as it is very difficult to use the personal data of the data subject in a way that is transparent and comprehensible, you should refrain completely from using such data in order to avoid the GDPR being applicable.

Copyright 

The question that arises under copyright law is: when employees use AI are they nevertheless originators of the created material? According to Section 7 of the German Copyright Act, this would be the case if it was obvious that the employee had to be designated as the creator of the work. From this point of view, AI tools thus likewise have to be used as auxiliary tools and the user always has to perform a final check. 

Please note: In particular, if the employer undertakes to provide an exclusive product for a customer then the employer could make itself liable on account of the lack of copyright protection.

Guidelines /operational instructions for employees 

From the aforementioned legal problems of a nature pertaining to employment rights, data protection legislation and copyright, it is clear that AI tools should be understood to be auxiliary tools. During such use, care should be taken to ensure that only data is used that is abstract, fully anonymised and non-personal and that the user always performs a final check. 

According to the current interpretation of employment law, AI tools are provided for in the right to give instructions in Section 106 GewO; consequently, this gives employers the possibility to control the use of AI via guidelines and operational instructions. It would be advisable to actually make use of this possibility and to take into account the above-mentioned aspects in the instructions. 

Recommendation: In this way, employers would ensure that there are no breaches of data protection laws or of copyright at their companies. Furthermore, employers would put themselves in a favourable position in terms of employment law as, from the existence of such a guideline, every time employees’ behaviour runs counter to it the employers would then be able to establish that there has been a breach of duty.

Outlook: If the rules for dealing with AI are addressed at an early stage in the form of a guideline then employees can be made aware of how to deal with AI in the long term and be optimally prepared for all further future technical developments.

Back
Back to top of page