Kanzlei OsujiLaw firm
Munich

Employment Law · AI Law

AI in human resources: legal considerations for employers in 2026

AI in recruitment, workforce planning and performance evaluation requires a review under the GDPR, works constitution law and the EU AI Act.

Read the German original

Laptop with a dark screen beside a stack of application folders and a pair of glasses on a desk.
This image is AI-generated.

Current case law: Co-determination rights of the works council when AI is used.

Automatically pre-sorting CVs, forecasting staffing needs or analysing employee performance: artificial intelligence in HR promises efficiency. Legally, however, even the choice of system, its introduction and the processing of employee data can have considerable consequences. Managing directors, HR managers and works councils should therefore clarify at an early stage which functions are permitted and where co-determination or additional protective measures become necessary.

Where is AI used in HR?

Typical fields of use are recruitment and applicant management (screening CVs, pre-selection and organising interviews), HR administration (duty rosters, internal chatbots, draft contracts), personnel development (training recommendations) and performance and behaviour analysis. The legal assessment does not depend on the product name alone, but on the specific purpose and use.

Everyday generative AI also carries risks: anyone who enters applications, health data or confidential company information into external tools without checking can breach data protection and confidentiality obligations. An internal AI usage policy should therefore set out specifically the permitted systems, input data, access rights, checking of results and responsibilities.

EU AI Regulation: When is recruiting AI a high-risk system?

The EU AI Act classifies certain AI systems for employment and personnel management as high-risk under Annex III No. 4. This may include systems that filter applications or evaluate applicants, take part in decisions on promotions or the termination of employment relationships, or assess employee performance and behaviour. Classification always requires an examination of the statutory conditions and possible exceptions. (EU AI Regulation (AI Act))

Important timetable: Under the amending Regulation (EU) 2026/1744, the special provisions for stand-alone high-risk AI systems under Annex III will in principle apply only from 2 December 2027. Other parts of the AI Act already apply. Employers must not mistake the later date for a general release for uncontrolled AI applications: the GDPR, the AGG and the BetrVG apply independently of it.

What is already prohibited in the workplace?

Article 5(1)(f) of the AI Act prohibits in principle AI systems that infer emotions in the workplace from biometric data. Exceptions are narrowly limited to medical or safety purposes. An ordinary fatigue, productivity or satisfaction analysis can therefore not be brought under an exception across the board. Not every system for analysing text or work performance is biometric emotion recognition; the technical function is decisive.

When does the works council have co-determination rights on AI?

Under section 87(1) No. 6 BetrVG, there is a right of co-determination when technical devices intended to monitor the behaviour or performance of employees are introduced and used. Objective suitability for monitoring may already be relevant. Careful examination is therefore needed for AI tools with performance profiles, activity analyses or analyses of communication data. (Section 87 (1) No. 6 of the BetrVG)

For selection guidelines, the co-determination rules under section 95(2a) BetrVG expressly apply even where AI is used to draw up the guidelines. In addition, rights to information and consultation under section 90 BetrVG and, for substantial changes to the establishment, sections 111 ff. BetrVG may become relevant. Not every AI tool automatically triggers all co-determination rights. (Section 95 (2a) of the BetrVG)

For larger companies, a framework works agreement on AI is advisable. It would set out the introduction process, areas of use, data access, limits on control, human review and regular evaluation. It does not replace the review of the specific system.

What data protection obligations apply to AI in HR?

The processing of personal data in the employment context needs a sound legal basis. In particular, purpose limitation, data minimisation, transparency, access rights, deletion and security of processing under the GDPR must be reviewed. Additional requirements apply to sensitive data such as health data.

A data protection impact assessment under Article 35 GDPR is required if the planned processing is likely to result in a high risk to the rights and freedoms of natural persons. It is not automatically mandatory for every AI tool. Likewise, Article 22 GDPR must be examined separately for automated individual decisions with legal or similarly significant effects.

Companies should also clarify where inputs and results are stored, whether an external provider uses data for its own training, which data processing agreement and third-country rules apply, and how data subjects’ rights are fulfilled. (Data protection law for companies in Munich)

Checklist: introducing AI in a legally secure way

  1. Document purpose and use: what is the system specifically meant to do, and which employees are affected?
  2. Classify the AI system: distinguish prohibitions, possible high-risk classification, and current and future AI Act obligations.
  3. Check data protection: legal basis, data flows, deletion periods, data processing agreements and, where necessary, an impact assessment.
  4. Involve the works council early: check participation rights against the specific application and shape agreements.
  5. Ensure human oversight: be able to trace AI outputs, monitor errors and discrimination risks, assign responsibilities.
  6. Introduce an AI usage policy: set rules for permitted tools, data categories, approvals, training and regular review.

Frequently asked questions about AI in employment law

May employers assess applications automatically with AI?

In principle, AI applications can be permitted. However, they must comply with the applicable data protection, equal treatment and co-determination rules. Depending on its function, the system may be classified as high-risk AI under Annex III.

Does the works council have to agree with ChatGPT or other AI tools?

Not merely because a tool uses AI. What matters are its functions, introduction and use, in particular possible monitoring of performance or behaviour under section 87(1) No. 6 BetrVG.

Does the AI Act already apply in 2026?

Yes, various provisions already apply. Under the revised timetable, the specific high-risk obligations under Annex III will in principle only apply from 2 December 2027.

Is AI-based employee monitoring allowed?

It can be problematic under data protection law and works constitution law. Biometric emotion recognition in the workplace is prohibited in principle. Other monitoring functions require separate examination.

Conclusion: AI in HR needs clear legal guidelines

AI can considerably support recruitment and personnel administration. For employers, however, it is crucial to shape technology, purpose, data protection and co-determination together. It is particularly risky to introduce new systems without first reviewing them and to clarify the legal questions only after a conflict.

Kanzlei Osuji advises employers on employment law questions about introducing new technologies and on employee data protection. A confidential enquiry is also possible by telephone on +49 174 343 0142. (Request an introductory call)

AI law and data protection

Labour law implementation

Author: Noah Osuji, lawyer. As of: 7 November 2026. General legal information, not individual legal advice.

Check AI errors, qualification and workforce reduction separately

Using a system does not yet answer the question of what duties employees have in individual cases and what training is required. Errors, unauthorised use and lack of skills are different starting points. The guide on AI use and dismissal explains duties to check, company rules and less severe measures. If the company is planning job cuts, the questions of reconciliation of interests, social plan and continued employment are added.

Next step

We agree the scope of work and fees before you instruct me.