Artificial intelligence is regarded as a key technology for the coming years. Yet the more sensitive the areas of application become – for example, in critical infrastructure, the public sector or organisations dealing with security – the more pressing another question becomes: just how sovereign is the use of AI? In conversation with AI experts Thomas Feld and Carsten Seck from Materna, it quickly becomes clear that sovereignty is not merely a technical label, but a strategic stance – and often the decisive factor in determining whether AI can be used productively or not.
Anyone looking for a clear definition will be met with a setback right from the start. “There is no single, fixed concept of sovereignty,” Thomas Feld makes clear right at the outset. Sovereignty can neither be attributed across the board nor assessed in binary terms.
“You cannot say: this system is sovereign, that one is not.”
What is always decisive is the specific use case and the requirements of the organisation wishing to deploy AI.
Rather than an ‘either/or’ scenario, the panellists describe a spectrum of sovereignty. At one end are fully open-source-based on-premises solutions offering maximum control; at the other, public cloud offerings from major hyperscalers, which can also be operated with sovereignty under certain conditions. In between, there are numerous gradations. Carsten Seck sums it up as follows:
Sovereignty means “having access to and the ability to intervene in the systems used in accordance with a pre-defined set of rules” – and thus being able to manage security, data protection and data processing in a transparent manner.
The extent to which these controls are set is a conscious decision.
Thomas Feld adds a strategic perspective to this view. For him, the focus is less on individual solutions and more on organisations’ freedom of action. “By sovereignty, we mean the sovereignty of companies and public authorities,” he explains. Above all, this refers to the freedom to make decisions without being constrained by technological, economic or geopolitical dependencies.
In his view, data sovereignty is an essential part of this. “Being in control of my own data” is a key component. Equally important is economic independence. Nobody wants to find themselves in a situation where their own business or key administrative processes only function as long as a single provider plays along. For state actors, there is an additional layer: political sovereignty – that is, the ability to define their own rules and laws governing the use of AI and new technologies.
It is precisely here that the difference between a sovereign and a purely technology-driven AI approach becomes apparent. A technology-driven approach is often geared towards the maximum range of functions or the greatest innovative edge. A sovereign approach, on the other hand, begins with the question: What are my requirements, and what am I prepared to accept in order to meet them?
“Technologically sovereign use of AI actually means choosing the right platform depending on my requirements,”
says Carsten Seck. Hyperscalers offer enormous technological advantages and global scalability, but inevitably lead to economic dependencies. National or European platforms are often less advanced, but in return they enable greater independence. Hybrid platforms, on the other hand, attempt to combine the best of both worlds.
The implication is clear: autonomy does not come from renunciation, but from conscious choices. Thomas Feld therefore emphasises the importance of multi-cloud strategies. Those who can work with different platforms and are not tied to a single provider increase their own freedom of choice and, with it, their autonomy.
Nevertheless, the productive use of AI in many organisations is falling short of expectations, particularly in KRITIS companies, the public sector, security-related organisations and the defence sector. According to Thomas Feld, the reason lies in a fundamental tension: on the one hand, there is a desire to utilise cutting-edge technology; on the other, there are extremely high requirements regarding security, data protection and compliance. According to the Federal Statistical Office, 51 per cent of companies cite a lack of clarity regarding legal implications as the reason for their reluctance to deploy AI systems productively. Achieving both at the same time is complex. The modernisation backlog can only be resolved through legally compliant AI.
By the end of the discussion, it becomes clear that sovereignty in the context of AI is not a technical detail, but a strategic guiding principle. It determines whether organisations can not only pilot AI, but also deploy it productively in a sustainable and responsible manner. Or, to put it another way: it is not the most advanced AI that is the best, but the one over which one retains control.
Materna demonstrates how sovereign AI can be implemented in practice through its six Sovereign AI Core Solutions. These cover key areas of application in which artificial intelligence effectively reduces the workload on organisations whilst simultaneously meeting high standards of security, transparency and compliance.
These include agent and assistant systems that support specialists in complex tasks or carry them out semi-autonomously, as well as AI-supported service and process automation, which enables end-to-end workflows to be automated efficiently and in compliance with regulations. In knowledge and document management, AI harnesses large volumes of unstructured information and makes it usable in a context-sensitive manner. Data analysis and forecasting systems make it possible to identify patterns, risks and trends at an early stage and to make well-informed decisions. The portfolio is complemented by digital twins, which synchronise the real and digital worlds for simulation and control, as well as digital situational overviews, which consolidate heterogeneous data sources into an up-to-date, reliable overall picture.
Together, these six Sovereign AI Core Solutions form the foundation for sovereign, practical and scalable AI deployment – tailored to the specific requirements of businesses and public authorities.
In public authorities and KRITIS organisations, AI-based agent and assistance systems are deployed where specialist staff are tied up by heavy documentation and data workloads. Real-world projects demonstrate how AI explains specialist procedures, structures files and consolidates information from various systems. The systems operate within the customer’s existing infrastructure and significantly reduce the manual research and copy-and-paste workload for staff.
AI-supported process automation is used in particular for complex, heavily regulated workflows. Examples range from agent-based review and approval platforms and applicant management solutions to Law2Logic, where changes to legislation are automatically analysed and directly incorporated into decision-making logic.
In knowledge and document management, AI helps organisations make use of large volumes of unstructured information. In practice, this involves, for example, the automated structuring of legal files, the generation of product and technical documentation, or the creation of AI-supported knowledge workstations for electronic files.
Digital situational awareness systems are used wherever decisions need to be made under time pressure based on fragmented data. Practical examples include shared situational awareness systems for the federal and state governments, risk and crisis management in critical infrastructure, and local resilience projects for towns and municipalities. AI automatically consolidates data from sensors, GIS systems, reports or other sources, assesses the overall situation and highlights critical developments for decision-makers.
Digital twins are used to continuously synchronise real-world infrastructure and processes with digital models. In projects such as the sustainable management of a state-owned forest, risk management for energy and gas networks, or municipal resilience initiatives, they enable simulation, analysis and predictive control. Discrepancies between the digital model and reality are identified at an early stage, enabling bottlenecks, wear and tear, or safety risks to be addressed preventively.
AI-supported analysis and forecasting systems help organisations derive reliable recommendations for action from big data at an early stage. Examples include strategic transport planning, visitor flow analyses, predictive maintenance and the monitoring of critical infrastructure. Forecasts help to identify risks and bottlenecks before they have an operational impact.
Regulatory requirements are the key driver behind Materna’s Sovereign AI Core Solutions. AI-powered systems such as video surveillance and property monitoring help customers to fulfil regulatory reporting obligations automatically and in compliance with regulations, particularly in the areas of digital situational awareness and digital twins. AI models are also used for fraud detection, for example to identify false identities or manipulated photos on platforms. Solutions such as Law2Logic automatically analyse changes to legislation and incorporate relevant updates directly into decision-making logic. This enables regulatory authorities to update specialist procedures, text modules and communications in a matter of days rather than months.
