Brandenburg's LLMoin AI Rollout Stirs Union Backlash as EU AI Act Enforcement Nears

Berlin-Brandenburg unions push for binding AI rights ahead of LLMoin rollout, while EU AI Act enforcement on August 2 brings stiff penalties and new compliance risks.

Unions in Berlin-Brandenburg are demanding binding co-determination rights as the region prepares to deploy "LLMoin," an AI language model for its state administration. During a gathering in Oranienburg on Friday, representatives from the DGB Berlin-Brandenburg criticized that works councils and staff councils have not been sufficiently involved in the planning process. Ver.di is pushing for mandatory service agreements at the state level, grounded in the Landespersonalvertretungsgesetz (State Staff Representation Act), to create guidelines for the public sector that go beyond mere information.

The push for worker input arrives as a broader surge in AI adoption reshapes German workplaces. A recent Ifo survey found that 54.5 percent of companies now use AI solutions, up sharply from 40.9 percent a year earlier. Adoption is highest in manufacturing (58.7 percent) and services (56.2 percent). Parallel research from SD Worx indicates that 48 percent of HR managers are actively investing in AI applications, and 45 percent of firms have already altered their workflows because of the technology.

In the private sector, the legal foundation for co-determination is already clearer. Under the Betriebsverfassungsgesetz (Works Constitution Act), works councils have a say when AI systems can be used to monitor employee performance or behavior. Employers must inform the worker representation body in good time about plans for technical equipment and work procedures.

Advertisement

Just as AI governance requires structured documentation and employer accountability, UK businesses face equally binding duties under workplace safety law. A free Health & Safety Toolkit provides ready-to-use risk assessments, checklists, and toolbox talks to help you meet your obligations under the Health & Safety at Work Act and related regulations — all instantly downloadable. Download the free Health & Safety Toolkit

Countdown to August 2: Stiff Penalties for Non-Compliance

The regulatory landscape will tighten considerably this summer. On August 2, Article 50 of the EU AI Act takes full effect, imposing strict labeling requirements for AI-generated content. That includes deepfakes, synthetic media, chatbots, and even photorealistic product images created with AI — all must carry an "AI-generated" mark. The same day, Article 4 becomes enforceable, mandating formal AI training programs to ensure workforce AI competence. Penalties can reach up to €35 million.

Combined with sanctions under the GDPR and the NIS2 Directive, cumulative fines could in extreme cases total up to 13 percent of a company's global annual revenue. Company directors also face personal liability for ensuring AI governance structures are in place.

Shadow AI and Systemic Security Gaps

Legal risks are only part of the challenge. Reports warn of "shadow AI" — employees using third-party tools without approval processes, leading to uncontrolled data flows and new cybersecurity threats. Faulty or "hallucinating" AI systems can cause damage comparable to classic cyberattacks.

The latest AIRQ Report on AI agents in enterprises underscores the danger: only 11 percent of systems studied had a robust security status. The majority possess a critical damage radius, with applications accessing private data and capable of autonomous actions.

Advertisement

Facing penalties and personal liability risks from non-compliance is a reality across all regulatory areas — health and safety included. A free toolkit offers nine practical tools including risk assessments, checklists, and a directors' liability guide designed specifically for the Health & Safety at Work Act 1974. Download the free Health & Safety at Work Act 1974 Toolkit

Experts recommend specific AI works agreements that address both EU AI Act requirements and employee data protection under the GDPR. A legally sound AI rollout, they say, requires close coordination among IT security, data protection officers, and worker representatives.