Language:
Woman sitting on the couch with a tablet

Trustworthy AI

Security for digital public administration

Artificial Intelligence (AI) holds enormous potential for the public sector. But if it is to be used with confidence, AI systems need to be trustworthy. Bundesdruckerei has set out clear principles for this.

AI in public administration: Great opportunities, significant concerns 

For public administration, the use of Artificial Intelligence promises major benefits: AI systems can automate processes, improve data-driven decision-making and make workflows more efficient, for example when handling applications. At the same time, the concerns are substantial: There are question marks surrounding data protection and the transparency of AI responses. Trustworthy AI is therefore the key to bridging the gap between technical feasibility and societal acceptance. The Bundesdruckerei-Gruppe is working intensively on the definition and significance of trustworthy AI and is exploring its practical implementation across various projects.

Four dimensions of trustworthy AI

The Bundesdruckerei-Gruppe defines the trustworthiness of AI on the basis of four core principles: Accuracy, explainability, data security and sovereignty. These principles provide important guidance for our work in AI.

 

Ansicht des Covers vom Buch Künstliche Intelligenz und Wir
Artificial intelligence and us

The book provides a sound overview of the current state of technological development as well as the future prospects of Artificial Intelligence (AI). From page 151 onwards, Carmen Dencker and Kim Nguyen (Innovations, Bundesdruckerei GmbH, Berlin) summarise the key points on trustworthy AI.

free download

Accuracy

Trustworthy AI systems do not invent or speculate; they adhere strictly to facts. They rule out hallucinations and false statements, and highlight contradictions in sources. This enables decisions in public administration to be made in a legally compliant, fair and factually correct manner, for example when processing applications or approvals.

Explainability

Trustworthy AI solutions make it clear where information comes from and how decisions are reached. This makes it clear why particular sources were selected and the legal basis on which a response rests, for example a piece of legislation. This enables staff in public administration, as well as citizens, to understand automated processes. It also enables them to check whether the results are trustworthy.

Data security 

All data, from source data to prompts and feedback loops, are protected throughout the entire data life cycle. Access rights and security classifications must be enforced at every stage of processing. This results in the highest security standards, with control over all data and the associated infrastructure remaining in the hands of public administration.

Sovereignty

Trustworthy AI should prevent lock-in effects and promote the development of in-house AI capabilities. In this way, flexible solutions can safeguard users’ long-term operational capability and independence.

Trustworthy AI products and projects

These principles for trustworthy AI are applied, for example, at the AI Competence Centre, where Bundesdruckerei GmbH pilots and promotes the use of AI in the federal administration. To this end, the KI-KC identifies suitable use cases, tests models and develops prototypes. Assistent iQ shows what such a solution can look like in practice. The text assistant, tailored to public administration, can search within documents, summarise them and assist with drafting.

Mann und Frau am Schreibtisch mit kleinem Roboter
AI Competence Centre

The AI Competence Centre (KI-KC) tests and promotes the use of AI in the German Federal Administration.Bundesdruckerei is working with public authorities to test AI use cases and develop user-centred AI prototypes in order to advance the digital transformation with the help of AI.

Find the project page PLAIN here
PLAIN: a data analysis platform for the entire German federal administration

The public administration works with large amounts of data and has started to use artificial intelligence (AI). In order to process this data securely, the federal administration needs robust and efficient tools.

Focus topic: Trustworthy Artificial Intelligence