Public Sector Must Ensure Human Rights Compliance in AI Use

The German government has issued guidelines to govern the responsible and secure deployment of artificial intelligence (AI) within federal administration. These guidelines aim to provide a framework for ensuring that all AI applications are transparent and compliant with fundamental rights.

In response to rapid technological advancements and increasing interest in the safe use of generative AI models such as ChatGPT, Gemini, and Claude, the Federal Ministry of the Interior (BMI) has formulated and published these essential guidelines. The recommendations are intended to serve as a guide for public sector employees to ensure that AI is integrated and utilized in accordance with constitutional rights.

One of the core principles highlighted in the guidelines is that AI must be deployed in alignment with fundamental rights, particularly regarding the privacy and informational self-determination of citizens and employees. The goal is to expand human agency through the use of AI, while ensuring that the technology is applied fairly and without discrimination based on characteristics such as gender, sexual orientation, disability, or age. AI systems should rely on suitable and representative training data tailored to their specific purposes, with a general rule that the most restrictive case of data input applies.

Furthermore, the guidelines stress that there must be clear accountability when employing AI technologies, in line with legal frameworks such as the AI Act. The use of AI should occur under adequate human oversight, with appropriateness assessed on a case-by-case basis. A challenging aspect of these guidelines is the requirement to trace and evaluate every outcome based on all underlying factors, a principle that many AI system operators may find difficult to adhere to.

Maximizing benefits while minimizing risks is another key focus of the document. It advocates for a shared values-based approach to the provision and utilization of AI, aiming to amplify the advantages of AI while mitigating associated risks. Applications should always be directed towards the common good, and public authorities are encouraged to deploy AI where it provides added value to their responsibilities.

The guidelines also touch upon sustainability, asserting that AI systems and their applications should be as resource-efficient and energy-conserving as possible. Efficiency considerations should be integrated into the development and procurement processes, with an emphasis on minimizing energy consumption and material resources required for AI computing power. Additionally, the technology can be actively utilized for initiatives that promote ecological sustainability. Factors such as the potential for future transitions, the agency of authorities over IT providers, and their market positions should be considered in order to avoid 'lock-in effects' and maintain digital sovereignty. There is also a recommendation to prioritize models with transparent training processes and openly accessible parameters.