Recruiting AI Gets Clearer Rules as EU Eases Companies' AI Competence Duty
04.07.2026 - 05:38:31 | boerse-global.de
European employers grappling with the integration of artificial intelligence into hiring will face lighter competence obligations under a finalised regulatory package, even as separate transparency mandates ramp up from August 2026. The “Digital Omnibus on AI”, adopted by the Council of the European Union on 29 June 2026, reworks key provisions of the bloc’s AI Act and introduces staggered deadlines that give many firms more breathing room.
Softened duty, preserved records
The most significant shift concerns Article 4 of the AI Act, which has been in force since February 2025. Originally, organisations were required to “ensure” a specified level of AI competence among their employees — a strict outcome-based obligation. Under the Omnibus, that has been replaced with a duty to “support” the development of AI skills. The change, described by practitioners as a shift from an “ensure” to an “endeavour” standard, removes the threat of strict-liability penalties for failing to achieve a predetermined competence level.
However, employers must still document concrete measures taken to foster skills. The German regulator — the Federal Network Agency (Bundesnetzagentur) — can inspect compliance from August 2026.
As employers prepare for regulatory scrutiny of AI competence measures, documenting workplace risks in traditional areas remains equally essential. A free toolkit with 41 ready-to-use templates helps you keep risk assessments compliant and up to date. Download the free Risk Assessment Toolkit
The revision responds to a persistent bottleneck. According to a survey by the digital association Bitkom, 53 percent of German companies named insufficient AI competence as the single greatest obstacle to adoption. While the AI Act theoretically allows fines of up to €35 million or 7 percent of annual worldwide turnover, the competence requirement itself carries no stand-alone penalty in the current framework.
Deadlines staggered for high-risk systems
For high-risk AI systems classified under Annex III — including software used for résumé screening, candidate ranking, or performance evaluation in HR — the conformity obligation now takes effect on 2 December 2027. Systems that function as safety components of already-regulated products (listed in Annex I) must comply by 2 August 2028.
National enforcement begins earlier: from 2 August 2026, the Bundesnetzagentur assumes market-surveillance responsibilities. The same date triggers broad transparency duties under Article 50: operators must label chatbot interactions, disclose any use of emotion recognition or biometric categorisation, and ensure these disclosures are clear to users. A transitional arrangement gives providers of existing AI systems until 2 December 2026 to adapt.
Recruiting: a practical test case
The implications for hiring are already being mapped out. A whitepaper published by the Queb network, which represents recruitment professionals, proposes a risk-traffic-light system to help companies classify their AI use in hiring.
- Green (low risk): purely administrative support tools.
- Red (high risk): any system that evaluates candidates or makes preselection decisions.
When a tool crosses into high-risk territory, the obligations become demanding: HR departments must maintain an explicit decision-making logic, create a traceable “audit trail”, and ensure meaningful human oversight. The first step, Queb stresses, is determining whether the tool qualifies as an AI system under the regulation at all.
The relevance is underscored by data from recruitment firm Michael Page: 67 percent of job-seekers in Germany already use generative AI for applications, while 55 percent of employers deploy AI tools during their selection processes.
While new AI governance frameworks demand attention, established workplace safety obligations remain in force — and over 37,000 UK companies already rely on a free Risk Assessment Toolkit with professional templates covering fire safety, manual handling, lone working and more. Download the free Risk Assessment Toolkit
Liability rulings and new governance
Recent court decisions have sharpened the legal stakes. In May 2026, the Munich I Regional Court (LG München I) ruled on liability for AI-generated content. Separately, the Hamm Higher Regional Court (OLG Hamm) confirmed that errors made by a chatbot can be attributed to its operator.
Companies are responding with more formal governance structures. A report by the EQS Group notes a rising trend toward centralised AI system registers and classification workflows within German firms.
On 2 July 2026, the federal government launched a cross-departmental AI task force led by the Digital Ministry. Its goal: by the Digital Summit in November 2026, to present a comprehensive AI strategy and establish a dedicated AI Safety Institute — moves Berlin says are needed to maintain competitiveness against the United States and China.
