AI Act implications for HR

Penalties vary based on the risk level associated to the AI system and the role in the supply chain.

In setting the obligations and penalties for not complying with the AI Act, it does matter whether you are a provider, importer, deployer or implementer (user) of an AI system. In this article, we will focus on the users: recruiting agencies, in-house HR departments, or contingency recruiters.

Are all HR AI systems classified as high risk under the EU AI Act?

AI tools for HR are under the radar as they can have significant power to negatively impact people’s lives. There are two verticals that need much care:

1. 1. recruitment or selection

1. 1. decisions regarding the terms of work

High risk AI systems in recruiting

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 are considered high risk AI systems according to ANNEX III in the AI Act. Here are some examples of tasks a high-risk AI system might solve:

1. 1. Making sure your job advertisement is reaching the “right” potential candidates through targeting

1. 1. Automatically get a ranked list of potential candidates based on the job description and different filters set in advance

1. 1. Find candidates with an increased chance of response by applying a “hidden AI” filter like “more likely to respond” or “more likely to change jobs”

1. 1. Assign a candidate score to each candidate, based on different features like: likelihood of succeeding, attrition risk, intelligence, work ethic

1. 1. Candidate profiling during a live job interview, to assess the personality type, attitude or estimate their success potential

1. 1. Conducting job interviews; having an AI virtual assistant perform the interview, asking questions and evaluating the answers or different personality aspects

1. 1. Deciding which candidates to move forward with

High risk AI systems in HR decision making

AI systems intended to be used to make decisionsaffecting terms of work-related relationships, the promotion or termination of work-related contractual relationships, to allocate tasks based on individual behaviour or personal traits or characteristics or to monitor and evaluate the performance and behaviour of persons in such relationships. Here are some examples of tasks a high-risk AI system might solve:

1. 1. Make decisions on whose contracts to terminate or whom to promote (based on cost-performance analysis, or risk to leave the company anyway, or predicted satisfaction score)

1. 1. Identify the best suitable 3 people in a team to assign a certain task (based on previous performance, age, attitude, personality type)

1. 1. Evaluate employee’s performance with a gamification AI-led test

1. 1. Analyze the employee’s behavior (leaves, day’s off, social interactions) with the purpose to decide on the promotion or termination of the contract.

Is my recruitment AI system profiling people?

Besides the list in ANNEX III, note that all AI systems and tools performing candidate profiling are considered high-risk AI systems.

The AI Act does not leave “profiling” subject to interpretation. According to Article 4 (4) in the Regulation (EU) 2016/679, profiling is achieved by using personal data (including economic, physical, mental, behavioral) in order to predict different aspects concerning a person’s life such as: their performance at work, economic situation, health, personal preferences, interests, reliability, behavior, location or movements.

Collecting data on the behavior of users (post interactions, interests) or inferring other data like gender or age (from post comments) and use that information for creating profiles, automatically puts high-risk on the recruitment platforms you are pooling talent from.

How much would AI transparency cost in recruiting?

One obligation as a user of an AI systemfor recruiting will be transparency. You will have to inform the candidates you reach out to that they were selected using AI-powered filtering or otherwise risk up to €7.5 Million or 1.5% of Global Turnover fines.

Action plan for an HR manager or recruitment agency

To make sure you stay ahead of the curve and avoid fines, crisis situations and disruptions, here is what you should consider ahead:

1. Audit

Evaluate the tools you are using and the process flow. Identify those tools that include AI components. Make sure you understand what happens behind all filters you apply, as some functionalities are just an interface to AI algorithms under the hood. Think if you could explain why candidates are ranked as they are. Clearly specify the points in the workflow at which AI and human intervention are following.

2. Risk assessment

Make sure to correctly determine the risk category each AI tool falls into according to the AI Act. This is extremely important as everything falling under high-risk AI systems impose strict obligations, like transparency, explainability and bias mitigation strategies.

3. AI Literacy

In such a sensitive field, where fairness and equal opportunities are valid ethical concerns, education and awareness are all the more important for HR professionals. Hosting workshops and trainings on understanding the implications of the AI Act, the AI technology’s opportunities, risks and impact, on understanding how an AI system works, how to interpret results and on the technology’s limitations should equip your HR team with necessary best practices. According to the AI Act all professionals using AI tools must undergo an AI literacy training, so take this as an opportunity.

4. Compliance

Update the company’s policy to include the AI Act, make sure the vendors you collaborate with foster a transparent relationship and answer your questions or look for different software solutions that comply.

Get it touch for more information on how we can take that burden off your shoulders at ioanang@giraffa-analytics.