European Union Artificial Intelligence Act: implications for HR Technology

In April of 2024 I was asked to provide thoughts on a draft of the new European Parliament Artificial Intelligence Act (EU AI Act). The purpose of the act is to provide “the proposal for a regulation of the European Parliament and of the Council on laying down harmonized rules on Artificial Intelligence”.? On June 7th I received a copy of the act scheduled to be released later in 2024. I was also given permission from one of the authors to publish my thoughts about the act, which led to this article.

?

The EU AI Act is 459 pages long and full of technical terms and legal jargon. After skimming through the document, I decided to summarize implications of the act as it relates to the use of HR technology using the help of AI (the irony of using AI to evaluate an act regulating the use of AI not lost on me). The following is a mixture of my thoughts about the act combined with content generated by Chat GPT 4o using prompts that limited responses to interpretation of text found in the act itself.? The implications are divided into three areas: relatively easy to manage, potentially difficult, and unknown given broader social factors.

?

Easy: Implications that should be manageable for use of HR Technology.

Much of the EU AI Act provides guidelines that will be familiar to anyone who has experience using HR technology and who takes issues related to discrimination, inclusion, and security seriously. The act creates a High-Risk classification for any AI systems used in employment for recruitment, selection, and evaluation of workers. This includes using AI for decisions affecting work-related relationships, promotions, and terminations. These AI systems must be designed so they do not perpetuate historical patterns of discrimination in the recruitment process and the evaluation of workers, and prevent biases against certain age groups, genders, disabilities, racial and ethnic origins. Any AI systems classified as high-risk or otherwise used for monitoring and evaluating workers must also comply with data protection and privacy regulations. Systems that could undermine workers' privacy rights are subject to strict oversight.

Many companies are already familiar with the challenges associated with ensuring HR technology does not discriminate against different groups and making sure HR data is appropriately treated to protect employee and candidate safety and security. These challenges are not easy, but they are manageable. The use of AI algorithms in HR technology should not, by itself violate any of these guidelines provided solutions are designed and use in an appropriate manner.? Although one particularly challenging area could be ensuring large language models used by AI solutions do not discriminate against people based on their linguistic heritage .

?

Difficult: Implications that could pose significant barriers to use of HR Technology.

The EU AI Act puts considerable emphasis on providing transparency and human oversight into use of AI algorithms. Companies must provide employees and candidates clear information on how AI is being used and how it affects decision-making processes that impact their careers and lives. Companies are expected to ensure human oversight to prevent automated systems from making critical employment decisions without human intervention. Providers of high-risk AI systems are required to have a quality management system in place to ensure compliance with regulatory standards. This involves systematic documentation of policies, procedures, and instructions related to AI system design and operation. Companies must conduct regular risk assessments and implement risk management strategies for AI systems used in HR. This includes identifying potential risks to workers' rights and implementing measures to mitigate them. Last, there must be mechanisms for taking swift corrective actions if an AI system fails to comply with regulatory standards.

The emphasis the EU AI Act places on ensuring humans are informed and involved in the application of AI technology is a notable departure from historic applications of HR technology where priority was often placed on using automation to eliminate human involvement to create cost savings. The focus on providing clear information to employee and candidates regarding the use of AI may require companies and HR technology providers to invest resources into creating mechanisms to communicate with people during talent processes such as recruiting, staffing, and compensation. This includes having processes that provide meaningful information about the use of AI in response to questions from candidates and employees. HR technology vendors may also have to build features that enable companies to conduct and act on the results of ongoing risks assessments into the use of AI.

?

Unknown: Implications that will depend on broader societal acceptance of AI.

The EU AI Act has significant implications for HR technology, but its focus is far broader spanning all aspects of society related to “the protection of natural persons, undertakings, democracy, the rule of law and environmental protection, while boosting employment and innovation”.? Much of the document delves into deep social values related to what is considered appropriate and inappropriate use of AU.? For example, the following uses of AI are specifically prohibited:

• Subliminal Manipulation: AI systems that deploy subliminal techniques to manipulate individuals’ behavior in a manner that works below the threshold of conscious perception
• Exploitation of Vulnerabilities: AI systems that exploit the vulnerabilities of specific groups of persons due to their age, physical or mental disability, or social or economic situation to materially distort their behavior in a manner that causes or is likely to cause them harm.
• Biometric Categorization: AI systems used for the purpose of categorizing individuals based on their biometric data, such as facial recognition systems used to infer or predict personal characteristics, that may lead to discriminatory outcomes.
• Emotion Recognition Systems: AI systems used to detect emotions or intentions of individuals in contexts related to employment, education, law enforcement, border control, and essential services.

The act also places significant limitations on use of AI related to law enforcement and military operations.?

The content in the EU AI Act about appropriate and inappropriate use of AI has implications for how companies can use AI to for job design, workforce management, and customer engagement.? It is hard to make specific predictions what this will mean for the use of AI in a work context in the near-term. What is clear is companies should carefully consider the provisions in the act when exploring use AI to redesign how employees do their jobs, including how they engage with customers and the broader market

Conclusion

Regardless of how one might feel about the EU AI Act, it will significantly impact the use of AI reated to HR technology and job redesign. ?Many provisions in the act will be readily familiar to experienced HR technology practitioners. But a lot of the content, much like AI itself, will take us into new and unfamiliar territory. Exploring new territory comes with risks, but these risks are often worth taking provided one recognizes and actively takes care to manage them.