EU Artificial Intelligence Rulebook Now Fully Enforced — Works Councils Scramble to Keep Pace
Published on 08/13/2026 at 16:03 | Redaktion boerse-global.de
The clock ran out on August 2, 2026. That was the day the European Union's Artificial Intelligence Act (Regulation 2024/1689) became fully applicable, ending a two-year rollout that began when the law first took effect on August 1, 2024. Core provisions around AI literacy and the most severe prohibitions had already been live since February 2, 2025, but companies now face the complete package of transparency and oversight duties.
The regulation draws a hard line on certain AI uses deemed incompatible with European values. Banned applications include manipulation techniques, social scoring, predictive policing, and emotion recognition in the workplace. The penalties for crossing that line are severe: fines can reach €35 million or 7% of a company's global annual turnover, whichever is higher.
Article 50 of the regulation, which took effect alongside the August deadline, introduces transparency obligations that partially overlap with Article 22 of the General Data Protection Regulation (GDPR). Businesses must now clearly disclose when AI systems are in operation. The only carve-out, per Article 50(4), applies when a human reviews the system's output. Ignoring these transparency duties carries fines of up to €15 million or 3% of annual turnover. Notably, the stricter requirements for high-risk systems have been pushed back by 16 months.
Works councils enter the picture
The European framework, however, does not directly dictate how AI gets deployed on the factory floor or in the office. That remains largely a matter for collective bargaining. Experts stress that works councils — the employee representation bodies common in German workplaces — still need to shape AI integration through company-level agreements. This is especially pressing in hybrid work settings, where transparency, human oversight, and active participation from employee representatives are essential.
As workplace technology evolves and oversight duties expand, making sure your own safety documentation is in order becomes even more critical. A free toolkit with 41 ready-to-use checklists and templates helps you document workplace risks properly and stay on top of your compliance duties. Download the free Risk Assessment Toolkit
The 2026 works council elections brought a wave of newly constituted bodies into office. These fresh faces now carry the responsibility of framing AI use in a legally sound way. Labour lawyers point to established case law from the Federal Labour Court, which underscores the inherent tension between employer and employee interests. Works councils, for instance, must approve overtime arrangements, and caps on working time accounts remain enforceable.
Adoption races ahead, skills lag behind
The regulatory urgency mirrors how quickly the technology has spread. Bitkom Research found that by March 2026, 41% of German companies with at least 20 employees were using artificial intelligence — up sharply from 17% a year earlier. In HR departments specifically, adoption sits at just 14%.
The technology itself is advancing at breakneck speed. By early 2026, AI agents could autonomously handle tasks lasting 14 hours, with performance reportedly doubling roughly every seven months. Yet obstacles persist. A BCG survey found that 55% of companies see redesigning role profiles as a major hurdle, and only 30% have dedicated staff working on AI governance.
Major vendors have moved to comply. Anthropic began watermarking Claude outputs worldwide on August 2, 2026, using invisible markers and C2PA metadata designed to survive copying — though not immune to format changes or translation. Industry heavyweights including Google, Microsoft, OpenAI, Meta, Aleph Alpha, and Mistral have also signed onto a voluntary code of conduct.
Guidance arrives for smaller players
Regulators have stepped in to help small and medium-sized enterprises and public administrations navigate the new rules. The data protection authority in Mecklenburg-Vorpommern (LfDI) published guidance on August 12, 2026, focused on low-risk AI applications like internal chatbots and document analysis.
With new regulatory layers landing on top of existing duties, many UK businesses are reviewing whether their safety documentation would stand up to scrutiny. A free Health & Safety Toolkit provides ready-to-use risk assessments and checklists covering key regulations like the Health & Safety at Work Act and COSHH — trusted by over 37,000 UK companies. Get the free Health & Safety Toolkit
North Rhine-Westphalia's labour ministry has issued a policy statement on AI in the workplace. Free basic training for apprentices is set to begin in September 2026, aiming to embed AI skills early in vocational education. On the technical side, Germany's Federal Office for Information Security (BSI) has released the "A5" assessment architecture, which takes both the EU AI Act and the Cyber Resilience Act into account.
