Quick HR Guide: Key considerations for Implementing AI to your German workforce

These days we are extremely busy advising clients on the implementation of AI tools and systems as well as navigating negotiations with works councils on that topic. 

This checklist shall give a brief overview on what to consider when implementing AI from a German employment law angle. Undisputedly, AI has a massive impact on the work environment and should be introduced in a conscious way. 

Alongside significant opportunities and game-changing technologies, there are potential risks, including the abusive use of AI, non-compliance with laws or intellectual property rights, and discriminatory decisions made by AI tools. Amongst others, Employers can be liable under the German General Equal Treatment Act if they have programmed or used AI systems inadequately, used an inadequate data basis or formula or do not carry out regular quality checks. 

For the liability it does not matter if the tools are external or internal and who (technically) was responsible for the error or discrimination practice. Additionally, total monitoring of employees’ is forbidden in Germany. Various further restrictions from the GDPR and the AI Act apply and must be considered. Depending on the concrete tools and setup, co-determination rights of the works councils must be observed and detailed works agreements are to be negotiated with the employee representatives.

To ensure things do not get out of control, companies must draft detailed and binding policies and guidelines. The following questions should be asked before any AI use is permitted:

• What tool shall be implemented and what data does it use?
  
• Who shall use the tool?
  
• What input? What output?
  
• For what is the tool used?
  
• Is there a works council?

Checklist AI Implementation

• Instruction or prohibition of AI use
  The employer can prohibit the use of AI as work equipment as well as instruct the use AI tools for fulfilling obligations within the scope of their right of direction under Section 106 GewO (observe any conflicting professional law, e.g. for doctors).
  
  
• Inventory
  Before creating binding regulations and compliance systems for the use of AI it is crucial to have a full and deep understanding of what tools, algorithms and systems there are in place, and which shall be implemented when. Already due to the AI Act it is required to undertake a solid inventory to ensure compliance with the obligations of companies using (or developing) AI.
  
  
• Process & Stakeholders
  A common process in practice is as follows: 

• Rights of the Economic Committee
  The Economic Committee might have the right to be informed about the introduction of AI tools as "new working methods".
  
  
• Rights of the Works Council
  If a Works Council exists, the use of AI tools can trigger information, consultation, and co-determination rights. Works councils themselves can consult experts and as a rule employee representatives must be taken along in the journey towards new technologies to ensure success. Transparency increases trust. For most technical tools, works councils must agree to the use and have a right to stop implementation before an agreement is found or replaced by a conciliation board decision. Data protection regulations must be observed!
  
  
• Introduction of rules of conduct regarding the use of AI applications
  Firstly, technical understanding of each tool is key. Thereupon, companies must decide what use and data input shall be allowed and how to deal with outputs. If no regulations exist yet, the personal, local, and material scope of application should be clarified first. It is also advisable to have clear rules on which AI tools may be used and to what extent. DP compliance and copyrights or trademark rights are to be ensured as well as restrictions of illegal use.
  
  
• Types of regulations
  Regulations on the use of AI tools can be made through works agreements and policies, in the employment contracts or amendments.
  
  
• Disclosure/labelling of use of AI tools and outputs
  Disclosure obligations of employees exist, and clear guidance is recommended, in particular regarding the intended handling, limitation of liability and corporate governance.
  
  
• Employee Monitoring
  Co-determination rights are to be considered. At the same time, it is essential to monitor legal and compliant use of AI systems and regulations in this regard.
  
  
• Protection of trade and business secrets and other confidential information as well as personal data
  When entering data that may contain business or trade secrets, it should be noted that some AI tools can use this information to improve their own services. Often, providers or third parties can access or learn with such data. It should be strictly regulated what kind of information employees are allowed to enter into the company’s chosen tools. This may require different regulations for different tools and often the one-size-fits-all approach is not advisable. 
  
  Sensitive data must also be protected from possible automated access by AI tools and DP breaches must be prevented at any times.
  
  This requires both technical restrictions as well as clear and binding regulations within the business.
  
  
• Input options for personal or confidential data
  See above: use should be strictly regulated and monitored. Awareness and technical understandings of data processing and borders are key.
  
  
• Liability
  Clear guidelines and corporate governance are key and control mechanisms to be established. Special rules apply for liability of employees in Germany.
  
  
• Prevention of discriminatory or non-compliant processes
  In particular, the employer should ensure through close monitoring and control systems that the AI has error-free data records and algorithms and creates outputs without discrimination or GDPR or IP rights concerns. Additionally, human supervision of each output is advisable.
  
  
• Control systems
  To ensure that only verified information and data is used, the employer should provide control systems. Monitoring and a proper compliance system must be established. Advisable is also to encourage employees to report incorrect or questionable results.
  
  
• Training and education
  Employers should offer their employees training courses or workshops, depending on their field of activity and the individual requirements of their job. The more sensitive the areas of activity (e.g. sensitive data), the more detailed the support for employees should be as well as training them about any risks and the correct use of AI tools!
  
  
• Limits to the use of AI tools in personnel administration
  Exclusively automated decisions ("profiling") are generally prohibited with few exemptions only; the final decision in matters with legal implications such as hiring, transfers, dismissals, etc. should therefore always be made and supervised by a human being.
  

Download the free PDF copy of the guide here