IAPQB

The EU AI Act and AI in hiring

What employers must do when AI helps to hire and manage staff, and what changed in 2026

IAPQB Report No. 8 · Rule explainer · Published 17 November 2026
International AI Practice Qualification Board (IAPQB)

In one minute

  • Since 2 February 2025, the EU AI Act has banned AI that infers emotions at work, and the European Commission says this includes hiring [1, 2]. A staff AI literacy duty applies from the same date [1, 3].
  • AI used to recruit people, decide on promotion or dismissal, allocate tasks or monitor performance is listed as high-risk [1]. The AI Omnibus, in force since 27 July 2026, moved these rules from 2 August 2026 to 2 December 2027 [1, 4, 5].
  • Employers using such tools must then follow the instructions for use, give oversight to trained staff, keep logs for at least six months and inform workers' representatives first. The Omnibus left these duties unchanged [1, 5].
  • German works council law already names AI [6]. In 2024 a Hamburg court refused a works council's request to stop staff using ChatGPT through private accounts [7, 8, 9].

1. A familiar situation

An HR manager in an EU company is asked to approve a video-interview tool that scores answers and reads emotions such as anxiety from candidates' faces and voices. The works council wants to see it first. The manager has read that the EU's rules for AI in hiring were delayed, and is unsure what applies today.

2. Why hiring tools get special rules

The AI Act treats AI in employment as high-risk because it "may have an appreciable impact on future career prospects, livelihoods of those persons and workers' rights" and may "perpetuate historical patterns of discrimination" [1, recital 57]. Annex III, point 4, covers AI that places targeted job advertisements, filters applications, evaluates candidates, decides on promotion or termination, allocates tasks based on behaviour or personal traits, or monitors performance and behaviour [1, Annex III]. Employers that use such tools are "deployers", and Article 26 sets their duties [1, Art. 26].

Emotion recognition at work is banned. The Commission ties this to "the imbalance of power" at work, and recital 44 notes "serious concerns about the scientific basis" of such systems [2, paras 241, 253].

IAPQB view. In the cases and guidance below, the answer turned on facts about one tool: whether it infers emotions, whether the employer can see what staff enter, and whether it only drafts text or also rates people [2, 8, 9, 10]. The 2026 changes moved the dates and left these questions in place [1, 5].

3. Three cases

3.1 Hamburg, 2024: a works council tries to stop ChatGPT

In December 2023, a medical technology company with about 1,600 staff at its Hamburg site allowed employees to use ChatGPT again, with guidelines, a handbook and an AI policy [7, 8, 9]. Staff used private accounts in a web browser, and the employer had no access to what they entered [8, 9]. On 16 January 2024 the Hamburg Labour Court rejected the group works council's urgent applications to stop the use [7, 8, 9]. It found that rules on using such tools concern how staff do their work, where the works council has no right of co-determination. It assumed that the provider recorded usage data, but the employer could not access it [8, 9].

3.2 Brussels, 2025: emotion recognition banned in hiring

Since 2 February 2025, the Act has banned AI systems "to infer emotions of a natural person in the areas of workplace and education institutions", except for medical or safety reasons [1, Art. 5, Art. 113]. The Commission's guidelines of 29 July 2025 say that "workplace" also covers candidates during selection and hiring. Their examples of prohibited uses include emotion recognition during recruitment or probation, and a call centre tracking staff emotions with webcams and voice recognition [2, para 254]. Fines for breaking a ban can reach EUR 35 million or 7% of a company's worldwide annual turnover, whichever is higher [2, para 55].

3.3 Strasbourg, 2026: the AI Omnibus moves the date

The high-risk rules for hiring tools were first due to apply on 2 August 2026 [5, recital 40]. Proposed by the Commission on 19 November 2025 and signed on 8 July 2026 as Regulation (EU) 2026/1744, the AI Omnibus entered into force on 27 July 2026 [4, 5]. Citing late standards and guidance and delays in setting up national authorities, it moved the rules for Annex III systems, including the deployer duties in Article 26, to 2 December 2027 [5, recital 40; 1, Art. 113]. Article 26 itself was not amended [1, Art. 26; 5].

Article 4 was amended. The Commission had proposed moving the AI literacy duty to the Member States and the Commission [3]. The final text keeps it on providers and deployers, who "shall take measures to support the development of AI literacy of their staff", with no required level [1, Art. 4]. Draft Commission guidelines on which systems are high-risk appeared on 19 May 2026 and were still a draft when this report was written [10, 11].

4. How practitioners handle it now

Organisations map which of their AI systems are high-risk. By December 2025, more than 230 AI Pact pledgers had committed to "mapping high-risk AI systems", and all had started internal processes to meet their commitments, the Commission reported [12]. Under the Commission's draft guidelines, a tool that only writes job descriptions from a recruiter's list of tasks and skills "should be exempt". One that sets the qualifications itself, or rates CVs and recommends candidates, "should be classified as high-risk" [10]. Emotion-reading features are already banned [1, Art. 5; 2].

Workers' representatives hear about a tool before it goes live, and national law can require more. Article 26(7) requires employers to inform workers' representatives and affected workers before using a high-risk system at work [1, Art. 26]. Article 2(11) lets Member States keep or introduce laws, and allow collective agreements, that protect workers more [1, Art. 2]. In Germany, employers must inform and consult the works council in good time on planned work procedures "including the use of artificial intelligence" (our translation) [6]. Selection guidelines for hiring, transfers and dismissals need its consent, also when drawn up with AI, and an expert counts as necessary when it must assess AI [6].

Oversight is assigned to people with training and authority. Deployers must assign it to people "who have the necessary competence, training and authority, as well as the necessary support" [1, Art. 26(2)]. The Commission says the duty to train staff for this oversight "remains in place" after the Omnibus [3]. Deployers must also keep logs under their control for at least six months unless other law provides otherwise, and tell people when a high-risk system makes or helps make decisions about them [1, Art. 26(6) and (11)].

Training is matched to staff and context, and no certificate is needed. Article 4 asks deployers to take account of staff knowledge, experience and the context of use [1, Art. 4]. The Commission says "There is no need for a certificate" and that organisations "can keep an internal record of trainings" [3]. AI Pact reports described "awareness campaigns, tailored training programmes and practical guidelines for everyday AI use" [12].

5. What to take from this

For HR staff who use AI tools. The guidance above points to two questions. Does it infer emotions? At work and in hiring, that is banned now [2]. Does it filter, rank or score people, or shape decisions on promotion, tasks or dismissal? Then it is probably high-risk, with duties from 2 December 2027 [1, 10].

For managers and employers. The Article 26 duties are known now, since the Omnibus left them unchanged [1, 5]. Briefing workers' representatives, training reviewers and storing logs take time to set up. German works council rights on AI already apply [6].

For works councils. German law treats an expert as necessary when a works council must assess AI [6]. In Hamburg, the court asked whether the employer itself could see staff data, and with private accounts it could not [8, 9]. The AI Act's duty to inform them about high-risk systems starts on 2 December 2027 [1].

6. Limits of this report

This report reflects official EU sources on 16 November 2026 and the consolidated AI Act text as at 27 July 2026 [1]. Commission guidelines are not binding [2, para 5; 10]. The Commission's AI literacy page dates the Article 4 change to "mid-July 2026"; this report uses the EUR-Lex date, 27 July 2026 [3, 5]. Only German national law was checked. The Hamburg ruling is a first-instance decision in urgent proceedings; its reasons come from two legal publications, and any appeal was not checked [7, 8, 9]. Data protection law is not covered, and no AI tool was tested.

References

[1] European Commission, AI Act Service Desk. [AI Act, Regulation (EU) 2024/1689; the article pages state that their text is based on the EUR-Lex consolidated version as at 27 July 2026.] Article 2: Scope, https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-2; Article 4: AI literacy, https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-4; Article 5: Prohibited AI practices, https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-5; Article 26: Obligations of deployers of high-risk AI systems, https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-26; Article 113: Entry into force and application, https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-113; ANNEX III, https://ai-act-service-desk.ec.europa.eu/en/ai-act/annex-3; Recital 57, https://ai-act-service-desk.ec.europa.eu/en/ai-act/recital-57 (all accessed 16 November 2026)

[2] European Commission (29 July 2025). Commission Guidelines on prohibited artificial intelligence practices established by Regulation (EU) 2024/1689 (AI Act). C(2025) 5052 final. https://ai-act-service-desk.ec.europa.eu/sites/default/files/2025-08/guidelines_on_prohibited_artificial_intelligence_practices_established_by_regulation_eu_20241689_ai_act_english_ied3r5nwo50xggpcfmwckm3nuc_112367-1.PDF

[3] European Commission. AI Literacy - Questions & Answers. Shaping Europe's digital future. https://digital-strategy.ec.europa.eu/en/faqs/ai-literacy-questions-answers (accessed 16 November 2026; page last updated 20 October 2026)

[4] European Commission (27 July 2026, last updated 31 July 2026). AI Omnibus enters into force. Shaping Europe's digital future. https://digital-strategy.ec.europa.eu/en/news/ai-omnibus-enters-force

[5] European Parliament and Council of the European Union (8 July 2026). Regulation (EU) 2026/1744 of the European Parliament and of the Council of 8 July 2026 amending Regulations (EU) 2024/1689, (EU) 2018/1139 and (EU) 2023/1230 as regards the simplification of the implementation of harmonised rules on artificial intelligence (Digital Omnibus on AI) (Text with EEA relevance). Official Journal of the European Union, L series, 24 July 2026. https://eur-lex.europa.eu/eli/reg/2026/1744/oj?locale=en (text read at https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=OJ:L_202601744)

[6] Germany. Betriebsverfassungsgesetz (Works Constitution Act), sections 80, 90 and 95. Gesetze im Internet. § 80 BetrVG - Einzelnorm, https://www.gesetze-im-internet.de/betrvg/__80.html; § 90 BetrVG - Einzelnorm, https://www.gesetze-im-internet.de/betrvg/__90.html; § 95 BetrVG - Einzelnorm, https://www.gesetze-im-internet.de/betrvg/__95.html (all accessed 16 November 2026)

[7] Arbeitsgericht Hamburg (16 January 2024). ArbG Hamburg Beschluss vom 16.01.2024 - 24 BVGa 1/24 [decision of the Hamburg Labour Court; operative part and part of the facts visible]. Haufe. https://www.haufe.de/id/entscheidung/arbg-hamburg-beschluss-vom-16012024-24-bvga-124-HI16327188.html

[8] Redaktion beck-aktuell (14 February 2024). Keine Mitbestimmung beim Einsatz von ChatGPT über Privataccounts. beck-aktuell. https://www.beck-aktuell.de/heute-im-recht/rechtsprechung/arbg-hamburg-keine-mitbestimmung-beim-einsatz-von-chatgpt-2024-02-14

[9] Betriebs-Berater (6 March 2024). Einstweiliger Rechts-schutz – Verbot des Einsatzes von ChaGPT und KI [editorial headnote and extracts from the reasons of ArbG Hamburg, 16 January 2024, 24 BVGa 1/24; full text BB-Online BBL2024-627-4]. Betriebs-Berater. https://betriebs-berater.ruw.de/arbeitsrecht/urteile/Einstweiliger-Rechts-schutz--Verbot-des-Einsatzes-von-ChaGPT-und-KI-45989

[10] European Commission (19 May 2026, last updated 23 July 2026). Draft Commission guidelines on the classification of high-risk AI systems. Shaping Europe's digital future. https://digital-strategy.ec.europa.eu/en/library/draft-guidelines-high-risk (general principles part: https://ec.europa.eu/newsroom/dae/redirection/document/128559; Annex III part, "Draft Commission guidelines on the classification of high-risk AI systems under Article 6 of Regulation (EU) 2024/1689 (AI Act) for stakeholder consultation": https://ec.europa.eu/newsroom/dae/redirection/document/128561)

[11] European Commission, AI Act Service Desk. Guidelines on the classification of high-risk AI systems. https://ai-act-service-desk.ec.europa.eu/en/guidelines-classification-high-risk-ai-systems (accessed 16 November 2026)

[12] European Commission (15 December 2025). AI Pact marks one year of progress on trustworthy AI in Europe. Shaping Europe's digital future. https://digital-strategy.ec.europa.eu/en/news/ai-pact-marks-one-year-progress-trustworthy-ai-europe

How to cite

IAPQB (2026). The EU AI Act and AI in hiring: what employers must do when AI helps to hire and manage staff, and what changed in 2026. IAPQB Report No. 8. International AI Practice Qualification Board. https://iapqb.org/research/008-eu-ai-act-hiring.html

About this report

Prepared by IAPQB. Each fact comes from the sources listed, and each source was opened and checked on 16 November 2026; anything that could not be checked was left out. AI tools were used to search and to draft, and IAPQB checked every claim against its source and is responsible for the text. IAPQB runs an AI work qualification; this report does not promote it. Corrections: contact@iapqb.org or an issue at github.com/iapqb/iapqb.github.io. Licence: CC BY 4.0.

Archived with the series on Zenodo: doi.org/10.5281/zenodo.24248199 · Markdown sources: github.com/iapqb/research