← Back to blog

AI

AI in recruiting: what the EU AI Act means concretely for mid-market employers in 2026

Published on 4 February 2026 · 5 min read

AI in recruiting is no longer a future topic. With the EU AI Act, it is also clear that specific uses fall into the high-risk category — with documentation and transparency obligations.

What the AI Act says

Annex III of the EU AI Act lists employment AI as high-risk — including systems for candidate selection, applicant assessment and promotion or dismissal decisions. Requirements include data quality, transparency, human oversight and conformity assessment.

For many SMEs the message is not 'switch AI off' but: document cleanly which systems are used in hiring, what training data they use and how a final human decision is guaranteed.

Where companies realistically stumble

Typical practical risks: an application portal filters CVs on criteria that act as proxies for age or origin. An AI interview tool rates speech patterns that systematically disadvantage non-native speakers. A matching algorithm is trained on historical hiring data that itself carries bias.

All of this is GDPR and anti-discrimination relevant and is additionally subject to structured documentation under the AI Act. Addressing this only after a complaint costs twice.

What this means for your business

In 2026, document every AI component in your hiring process, guarantee human final decisions and review screening criteria regularly for indirect discrimination.

Share on LinkedIn

The template text is copied to your clipboard on click.

Transparency, image credits and advertising disclosure

Responsible for editorial content pursuant to § 18(2) MStV: Sandra Lochmann, Bahnhofstraße 28, 82065 Baierbrunn, Germany.

Reference evidence: named quotations and client or partner references are published only with documented consent or a publicly verifiable source. Article-specific sources are listed below each post; business relationships are disclosed in the text.

Image credits: cover and preview images on this blog are AI-generated, editorially reviewed in-house productions by People · Places · Performance; usage rights are held by Sandra Lochmann. Portrait photography: © Sandra Lochmann, used with the consent of the person depicted. Third-party images, where used, are credited directly at the image with author and licence (e.g. Unsplash licence, CC BY 4.0). Third-party logos and trademarks are named for editorial reporting only; all rights remain with their owners.

Advertising disclosure: this article is editorial content and contains no paid advertising, no sponsorship and no affiliate links. Companies, regions, programmes and tools mentioned are named without commission or payment. Where a business relationship exists (e.g. ongoing mandates or partnerships), it is disclosed in the text. Paid cooperations would be clearly labelled 'Advertisement' at the beginning of the article (§ 5a(4) UWG, § 22 MStV).

AI notice: AI tools are used to support writing and translating these articles. Content, arguments, assessments and sources are editorially reviewed and taken responsibility for before publication (Art. 50 AI Act).