Starting August 2, 2026, any AI system that interacts with employees, screens candidates, or evaluates performance using generative or synthetic content has to disclose that it is AI, under Article 50 of the EU AI Act. The provision covers any employer using these tools “professionally or for business purposes,” and it applies to providers worldwide, not just companies headquartered in the EU. For HR teams that have quietly rolled out chatbots, AI-written policy communications, or synthetic training content without a compliance review, the deadline turns an operational shortcut into a legal one.
The rule requires providers to design AI systems so people know they are interacting with one, unless that is already obvious. Deployers have to label deepfakes clearly on first exposure and disclose the use of emotion recognition or biometric categorization. AI-generated or manipulated text published on matters of public interest, without human editorial review, also needs a label. National market surveillance authorities enforce it, with penalties reaching 15 million euros or 3% of global turnover, whichever is higher.
The compliance risk is broader than most HR functions have mapped. HRTech has covered how AI hiring and layoff tools are becoming employers’ biggest litigation exposure in the U.S.; Article 50 adds a parallel, disclosure-based obligation in the EU that does not require proof of discriminatory impact to trigger a penalty, only proof that a system generated or manipulated content without the required label. Any HR platform with EU-based employees or candidates should inventory every AI-facing tool now, from résumé screeners to internal chatbots, and confirm each one meets the disclosure bar before the grace period on marking obligations expires in December.
Source: European Commission
