The European Union is cracking down on AI use in the workplace. Starting December 2, 2027, companies must ensure that high-risk AI systems cannot make dismissal decisions autonomously – human review becomes mandatory. This requirement is part of the Digital Omnibus on AI, a regulatory framework drafted in November 2025 and formally adopted according to the source.
Key Facts
- December 2, 2027: Binding deadline for high-risk AI systems in personnel decisions
- August 2, 2028: Deadline for embedded AI systems (e.g., integrated into software packages)
- Human oversight: Dismissal decisions cannot be fully automated
- Phased rollout: EU AI regulation has been implemented in stages since February 2025
The EU's Regulatory Timeline
The EU AI Act does not take effect all at once – it follows a staggered schedule:
| Timeline | Regulatory Area | Status |
|---|---|---|
| February 2025 | Ban on unacceptable AI risks | In force |
| August 2025 | General Purpose AI obligations | In force |
| August 2026 | Transparency requirements | In force |
| December 2, 2027 | High-risk AI in personnel decisions | Upcoming |
| August 2, 2028 | Embedded high-risk systems | Upcoming |
This staggered approach gives companies time to adapt – but the requirements are complex.
What Counts as "High-Risk" in HR?
AI systems are classified as high-risk when they make decisions in sensitive life areas – and personnel management is one of them. Specifically: if a company uses AI to identify candidates for dismissal, that decision cannot be made by the algorithm alone. A human must review the recommendation and make the final call. This is not optional – it becomes legally binding in December 2027.
The rule aims to ensure that affected employees are not victims of algorithmic errors and that companies retain responsibility for their decisions.
Documentation Becomes a Compliance Hurdle
The new deadlines are only half the story. Companies must not only adapt their systems but also thoroughly document how they work and how human oversight is implemented. These documentation requirements are often underestimated and can lead to significant legal consequences if not met.
Companies using AI systems must be able to prove: What data feeds in? How does the system make decisions? Where does the human intervene? How is fairness ensured?
What This Means for German Companies
German employers using AI for personnel decisions – whether for recruiting, performance reviews, or dismissals – should act now. The December 2027 deadline is closer than it seems. Those without a compliance strategy should develop one: Which AI systems are in use? Are they classified as high-risk? How is human oversight implemented? What documentation is still missing?
With this regulation, the EU is sending a clear signal: AI in the workplace is no longer experimental territory – it is regulated terrain. Companies that react too late risk not only fines but also loss of trust among employees and candidates.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
