Nearly Half of German Public Agencies Already Using Generative AI, Survey Finds

A study by market researcher IDC, commissioned by Dell Technologies, found that 47.2 percent of government agencies are already using the technology, while another 50 percent plan to introduce it within the next 12 to 18 months. Yet the rapid adoption brings serious hurdles: 66.7 percent of respondents named data protection as the biggest obstacle, followed by regulatory uncertainty (55.6 percent) and a shortage of skilled staff (52.8 percent).

Despite those barriers, expectations are high. More than 80 percent of agency representatives see benefits for national security, over 60 percent anticipate productivity gains, and 86.1 percent believe so-called agentic AI will play a central role in the future. But the push for efficiency is colliding with legal limits designed to protect employees and citizens.

Co-determination rights for staff councils remain broad

German federal and state civil-service laws grant staff councils extensive co-determination rights when AI systems are introduced. That applies particularly when the technology can be used to monitor employees, reorganises work processes, or serves as a basis for hiring or performance guidelines. Labour law experts recommend that specific works agreements be signed. These should require strict purpose limitation, transparency about how algorithms function, and a “human-in-the-loop” principle that keeps a person as the final decision-maker. Mandatory training and regular evaluations are also deemed essential.

North Rhine-Westphalia imposes strict data-economy rules

One example of how states are tightening the rules comes from North Rhine-Westphalia. A guideline issued by the state’s digitalisation ministry under minister Ina Scharrenbach orders agencies to observe rigorous data economy. Transmitting personal data to AI systems is prohibited. Approved uses are limited mostly to text, images and language editing. The guideline explicitly bans applications that could manipulate the behaviour of citizens or employees — such as social scoring or automated performance monitoring — and stresses that human accountability cannot be replaced.

New digital tool targets rising mental-health risks

At the same time, a new digital stress matrix has been presented as part of an EU-wide campaign. It is designed to help staff and works councils identify risks early, focusing on areas such as task automation and AI-assisted personnel management. The need is pressing: in 2024, mental disorders accounted for a significant share of sick-leave days and early-retirement pensions in Germany. The ver.di trade union is calling for greater involvement of the public sector in strategic advisory bodies. In Rhineland-Palatinate, the organisation is pushing for a permanent seat on the state’s transformation council, arguing that digitalisation and AI should not be used solely as cost-cutting tools.

While German authorities tackle new risks from AI, UK employers face their own compliance pressures across workplace health and safety. Over 37,000 British companies already use a free toolkit to keep up with risk assessments, fire safety, COSHH and other legal duties. Download the free Health & Safety Toolkit

AI Act deadlines approach, bringing new transparency rules

The legal framework is set to tighten further. From 2 August 2026, the European Union’s AI Act will impose initial transparency obligations. Stricter requirements will follow from 2 December 2027 for high-risk applications, including AI used to evaluate examination performance. Parallel initiatives are already under way: in June 2026, a pilot hub for agentic AI was launched with 18 municipalities participating, aiming to speed up administrative processes. In Baden-Württemberg, a dedicated AI ecosystem built on a sovereign German cloud infrastructure is scheduled to go live in July 2026, bundling applications for multilingual citizen services and AI-assisted workspaces.

en | boerse | 69618730 |