AI Act Forces Polish Companies to Audit Recruitment and HR Algorithms

HR departments in Poland and across the European Union are entering a new legal regime. Starting August 2, 2026, the AI Act's rules for high-risk artificial intelligence systems take effect, and tools used for CV screening, performance evaluation, and identifying candidates for promotion now fall into that category. That marks the end of a setup where the algorithm decided and a human simply clicked approve.

The regulation covers tools for CV screening, employee performance monitoring, and systems supporting decisions on promotions and layoffs. All three categories were classified as high-risk because they directly affect access to employment, work evaluation, and employment conditions. The AI Act explicitly bans emotion analysis of employees, social scoring, and facial recognition in recruitment processes.

No More Rubber-Stamp Oversight

A key element of the new rules is the requirement for real, not merely declarative, human oversight of algorithmic decisions. Article 26 of the AI Act obliges companies to place that oversight in the hands of people with the right competence, training, authority, and organizational support. The regulation explicitly bars situations where an employee automatically approves a system-generated result without any real ability to challenge it.

The problem of pretend oversight shows up most often where a recruiter has to process hundreds of recommendations a day, has little time for each case, and is judged on speed. Under those conditions, the risk grows of what's known as automation bias, the tendency to trust a recommendation unquestioningly simply because a piece of technology produced it. The AI Act requires that the person providing oversight be able to ignore the system's output, replace it with their own judgment, reverse a decision based on an AI recommendation, and, where warranted, halt the system altogether.

What the Audit Covers

Under the new requirements, employers must conduct a full audit of the AI systems they use. That includes taking stock of the tools and vetting vendors, assessing risk in specific areas of the company's operations, implementing a risk management system with regular updates, and documenting processes and data-handling policies. On top of that comes a transparency obligation toward candidates and employees, who must be told that AI is involved in the process and given access to information about the factors shaping the system's recommendations.

Even though the new regulations don't take effect until August 2026, waiting until then to start preparing would be very risky - Katarzyna Szudy, HR Contact

Industry estimates put compliance costs anywhere from a few thousand to several hundred thousand euros, depending on the size of the organization and the number of AI tools in use. For large companies that already rely heavily on algorithms in recruitment, that means reviewing their entire HR technology stack before August 2026.

Poland's Compliance Gap

The data show a clear gap between large and small companies in Poland. While more than half of large enterprises already use AI in HR processes, the share among small companies doesn't exceed a few percent. That raises the question of whether smaller firms are even aware of the new audit obligations, and whether they'll be able to cover the cost of complying by the deadline.

The new rules fit into a broader regulatory trend around the AI Act in Poland, where companies and public institutions still struggle to determine which of the systems they use even fall under EU regulation. The August deadline for high-risk systems could therefore catch some employers unprepared, especially since full compliance for already-existing systems isn't required until August 2027, a date some companies may mistakenly read as extra breathing room.

For HR departments, this means changing in practice how they work with recruiters and managers who use algorithms. Simply deploying a tool from an outside vendor is no longer enough; companies must document who oversees its decisions and on what basis, and demonstrate that employees have a real, not just formal, ability to challenge the system's recommendations.