top of page

AI Act: What Applies in Business and What Is Coming

Nov 13
3 min read

The AI Act (KI-Verordnung) for businesses: obligations, deadlines and changes under the Digital Omnibus

According to media reports, the Federal Government is planning a partnership with technology and industrial companies to advance artificial intelligence in German factories. In many medium-sized businesses AI has long been part of everyday work – from text assistants to the pre-selection of applicants. Which rules already apply today, and what is still to come?

Legal Framework and Your Own Role

The European AI Regulation (Regulation (EU) 2024/1689; abbreviated KI-VO in German, in English the “AI Act”) applies directly. The so-called Digital Omnibus Regulation on AI (Regulation (EU) 2026/1744) amended it with effect from 27 July 2026: deadlines were postponed and individual obligations eased, while new prohibitions will be added from 2 December 2026.

What matters is your own role. A “provider” (Anbieter) is anyone who develops an AI system or has one developed and places it on the market or puts it into service under its own name – including free of charge, and therefore possibly already in the case of in-house development for internal use. A “deployer” (Betreiber) is anyone who uses an AI system under its own authority in a professional capacity (Art. 3 Nr. 3 and 4 KI-VO). Most businesses are deployers.

What Already Applies

Prohibitions: Since 2 February 2025 certain AI practices have been prohibited (Art. 5 KI-VO). Important for employers: AI systems that infer the emotions of employees in the workplace are in principle prohibited; exceptions exist only for medical or safety reasons (Art. 5 Abs. 1 lit. f KI-VO).

AI literacy: Providers and deployers must take measures to promote the AI literacy of their staff (Art. 4 KI-VO). Since the Omnibus it has been clarified that no particular level of competence has to be guaranteed. Training and clear internal rules nevertheless remain called for.

Transparency: Since 2 August 2026 the transparency obligations of Art. 50 KI-VO have applied. Providers must design systems that interact with people, such as chatbots, so that users recognise that they are communicating with an AI. Deployers must disclose when they use AI to generate deceptively realistic image, audio or video content (“deepfakes”). A general obligation to label every piece of AI-generated content, however, does not exist.

What Is Coming: High-Risk AI in HR

Strict obligations apply to high-risk AI. Under Annex III of the KI-VO this includes, among others, systems for the selection of applicants, for decisions on promotion or dismissal, and for performance evaluation. After the Omnibus these obligations apply only from 2 December 2027; for AI in regulated products such as machinery, from 2 August 2028 (Art. 113 KI-VO).

Deployers will then have to use such systems in accordance with the instructions for use, ensure human oversight by competent persons and, as employers, inform employee representatives and affected employees in advance (Art. 26 Abs. 1, 2 and 7 KI-VO). Anyone who places a high-risk system on the market under its own brand or substantially modifies it is regarded from that point as a provider itself (Art. 25 Abs. 1 KI-VO).

Works Council, Data Protection and Supervision

The works council (Betriebsrat) must be informed in good time of the planning of work procedures “including the use of artificial intelligence” (§ 90 Abs. 1 Nr. 3 BetrVG). If an AI tool is capable of monitoring behaviour or performance, a right of co-determination under § 87 Abs. 1 Nr. 6 BetrVG may come into consideration. In addition, the GDPR (DSGVO) remains applicable.

The competent supervisory authority in Germany is in principle the Federal Network Agency (Bundesnetzagentur) (§ 2 Abs. 1 KI-Marktüberwachungs- und Innovationsförderungsgesetz, in force since 29 July 2026). Infringements of the prohibitions can be fined up to EUR 35 million or 7 per cent of worldwide annual turnover, infringements of certain further obligations up to EUR 15 million or 3 per cent – in each case whichever is higher (Art. 99 Abs. 3 and 4 KI-VO). For small and medium-sized enterprises the lower amount applies (Art. 99 Abs. 6 KI-VO).

Practical Checklist

  • Inventory: which AI systems do we use, for what, and as provider or deployer?

  • Rule out prohibited applications, in particular emotion recognition in the workplace.

  • Promote AI literacy: training, usage policy, persons responsible.

  • Check whether transparency obligations apply (chatbots, deepfakes).

  • Record AI in HR and prepare for the high-risk obligations by December 2027.

  • Involve the works council and data protection early.

Conclusion

The postponed deadlines buy time, but not a pause: prohibitions, AI literacy and transparency obligations already apply. Anyone who records and organises their AI use now will also be prepared for the high-risk rules. Please do not hesitate to contact us if you have any questions.

Status: 18 October 2026

NIETZER Rechtsanwälte · Attorneys at Law (USA) · Heilbronn · www.unternehmensrecht.com

This article is for general information purposes and does not replace legal advice in an individual case.

bottom of page