New institute to assess risks of modern AI systems and help shape international standards
The German government is stepping up its activities in the field of artificial intelligence and establishing a new body to assess the opportunities and risks of modern AI systems. The National Security Council has decided to establish a specialised AI institute, which is to serve as a central centre of expertise for analysis, assessment and advice on high-performance AI models.
With this initiative, the German government is responding to the growing importance of generative AI for the economy, public administration, cybersecurity and critical infrastructure. The aim is to assess technological developments more quickly and to better evaluate potential impacts on security, society and digital resilience.
Launch as a virtual organisation
In an initial phase, the institute is to be established as a virtual structure. This will draw on the existing expertise and resources of the Federal Network Agency and the Federal Office for Information Security (BSI). The establishment of a permanent location is planned for a later date.
The lead ministries – the Federal Ministry of the Interior and the Federal Ministry for Digital Affairs – intend to press ahead with the establishment swiftly in view of the rapid pace of innovation in the AI sector.
Focus on security and trustworthiness
The planned institute will address two key areas. On the one hand, the focus will be on the technical protection of AI systems against attacks and manipulation. On the other hand, the aim is to investigate risks that may arise from the systems themselves.
This involves, for example, the reliability of AI applications, the quality of their results, and potential errors or unexpected behaviour. The institute is intended to analyse modern models, assess their capabilities and identify potential risks at an early stage.
National expertise for global technologies
Another priority is international exchange. The institute is to work closely with comparable institutions in other countries and support the development of common standards.
Many countries already have specialised institutions for AI security. Germany now intends to consolidate its activities in this area more effectively and embed them institutionally. Until now, such tasks have primarily been carried out by the BSI.
Support for policy and administration
According to Digital Minister Karsten Wildberger, the new institute is to be staffed by highly qualified specialists. Their task will be to test new AI models promptly, assess their capabilities and provide expert advice to policy-makers and public authorities.
The British AI Safety Institute, which already plays a central role in evaluating high-performance AI systems, is seen as a possible model.
Particular attention is being paid to the latest developments in the field of generative AI. New models are increasingly achieving capabilities that both open up economic opportunities and raise new security concerns.
AI Act increases regulatory pressure
In parallel with the establishment of the institute, European regulation is also gaining in importance. The Federal Network Agency is set to become the central supervisory authority for the implementation of the European AI Act in Germany.
Among other things, the regulation introduces risk categories for AI applications and defines extensive transparency and labelling requirements. Parts of the regulations will become binding from August 2026.
Companies, public authorities and providers of AI solutions must therefore increasingly adapt to new requirements in the areas of governance, documentation and risk management.
AI security becomes a strategic task
The decision highlights that artificial intelligence is increasingly viewed as a key technology relevant to security. The new institution is intended to bring technological expertise, risk assessment and regulatory requirements more closely together.
This is likely to be an important step for companies, operators of critical infrastructure and public institutions. After all, the more AI is integrated into operational processes, the more important independent assessment bodies, robust security standards and a thorough analysis of potential risks become.
The planned institute could thus become a central component of Germany’s AI and cybersecurity strategy, whilst simultaneously strengthening Germany’s position in the international debate on trustworthy and secure artificial intelligence.
