EU AI Act: What HR Needs to Know
The EU AI Act is set to come into force in February 2025, affecting any company that develops or deploys large language model tools, including employers in the UK. This new legislation aims to regulate the use of artificial intelligence within recruitment, HR, and worker management domains, categorizing these applications as high risk.
The Rise of AI in HR
The excitement surrounding advancements in AI technology over the past 18 months, coupled with a challenging economic environment, has led many HR departments to adopt AI solutions rapidly. These tools, such as chatbots, can effectively handle candidate queries, schedule interviews, and provide application updates in a more engaging manner than traditional methods.
AI can also enhance recruitment by scanning job boards, social networks, and internal databases to identify candidates with specific skills. Furthermore, it can screen and rank CVs more efficiently than human recruiters, streamlining the hiring process.
AI technologies extend beyond recruitment; they can automate onboarding processes, assist in training design, analyze internal surveys, and detect trends in employee sentiment. Surprisingly, they can even help mitigate unconscious bias in performance reviews.
Forthcoming Legislation
Previously, the deployment of AI tools was primarily governed by data protection principles outlined in Article 5 of the GDPR. These principles emphasize accuracy, fairness, transparency, and confidentiality, along with regulations concerning “automated decision-making” in Article 23.
However, as of February 2, 2025, the EU AI Act introduces a new regulatory framework, classifying AI activities in recruitment, HR, and worker management as high risk. This means that UK businesses deploying AI systems within the EU will need to comply with the Act, adding another layer of legislative oversight.
Risks and Responsibilities
The introduction of the EU AI Act prompts critical questions about the reliability of AI systems. Concerns arise about potential underperformance or erroneous outputs. If AI systems introduce bias rather than eliminate it, misinterpret candidate qualities, or inaccurately analyze data, the repercussions can be damaging for both employers and employees. This could lead to hiring mistakes, incorrect reprimands, and inappropriate advertising, potentially harming company culture and resulting in costly litigation.
High-Risk Activities Under the EU AI Act
The EU AI Act categorizes recruitment, HR, and worker management applications as high-risk, placing them alongside sectors like medical devices, autonomous vehicles, law enforcement, biometric identification, and critical infrastructure.
Organizations creating AI systems for these applications are classified as AI providers and must adhere to several regulatory obligations, including:
- Implementing a comprehensive risk management process
- Utilizing data sets that are fit for purpose and free from bias
- Maintaining technical documentation and human oversight
- Completing a conformity assessment, similar to a data protection impact assessment but focused on AI
- Registering the model in the official EU database of high-risk AI systems
- Monitoring and correcting performance and safety post-deployment
Even if organizations use existing models, such as open-source large language models, they must comply with these responsibilities. Deploying systems sourced from other providers also entails accountability for requirements, including purpose limitation, human oversight, and incident reporting.
The Importance of Good AI Governance
AI operates differently from traditional software, as it is inherently probabilistic. This characteristic allows it to handle real-world uncertainties more effectively but introduces uncertainty in its outputs. Thus, compliance with the EU AI Act becomes essential for managing this uncertainty effectively.
Achieving robust oversight of data, fostering transparency, and designing processes that align with both business goals and candidate welfare are crucial steps. This entails conducting thorough audits of existing AI systems, implementing rigorous design practices for new systems, and educating teams on the ethical deployment of AI.
By adopting proactive measures and establishing solid AI governance frameworks, HR departments can navigate the complexities of the EU AI Act and harness the potential of AI technology, transforming it into an asset for their organizations rather than a liability.
