Charité – Universitätsmedizin Berlin is moving artificial intelligence from individual experiments towards shared hospital infrastructure. Germany's largest university hospital has developed a central AI platform that will provide employees with controlled access to different AI models and applications across clinical care, research, education and administration.
The platform, called Clé, has completed its initial testing phase and is now being opened progressively to employees across the organisation. Charité says the infrastructure has been specifically developed for university medicine and allows AI to be operated under the hospital's own responsibility, with security and data protection built into the environment.
The approach is significant because hospitals are increasingly confronted with dozens of potential AI applications while still lacking a common infrastructure for deploying them safely. Rather than allowing individual departments to introduce separate systems with different interfaces, governance arrangements and data requirements, Charité is building a central environment through which AI models and services can be made available across the organisation.
From individual AI tools to shared infrastructure
Healthcare AI adoption has largely developed application by application. Radiology departments introduce imaging algorithms, administrative teams experiment with generative AI and researchers use large language models or machine-learning environments for individual projects. While this can accelerate experimentation, it can also create fragmented technology landscapes in which every new application raises separate questions about security, data protection, integration and responsibility.
Clé is designed to provide a common foundation instead. The platform can make different AI models available through one controlled environment and offers interfaces that allow additional applications and services to be connected. This means Charité does not have to build its AI strategy around a single model or supplier but can create an infrastructure capable of supporting different technologies for different purposes.
That distinction is becoming increasingly important as generative AI develops rapidly. The model considered most appropriate for a research task may not be suitable for processing administrative information or supporting a clinical workflow. A central platform can potentially allow an organisation to change or add models while maintaining common requirements around access, security and governance.
Charité emphasises that Clé is operated under its own responsibility and has been designed for secure and privacy-compliant use. For a university hospital working with sensitive patient and research information, retaining control over how AI services are provided is particularly important.
One platform across the hospital
The scope extends well beyond clinical AI. Charité intends the platform to support healthcare delivery, scientific research, teaching and administrative processes, making it an institutional AI infrastructure rather than a specialist clinical system.
Potential applications are therefore broad. Generative AI could support employees in working with text and information, researchers could use specialised models and interfaces, and new AI-supported workflows could eventually be integrated into operational or clinical processes. The platform's architecture allows different models and services to be added as requirements develop.
This also changes the implementation challenge. Once AI becomes available across an organisation, technical access is only one part of adoption. Employees need to understand which tools can be used for which tasks, what information may be processed, how outputs should be checked and where professional responsibility remains with the user.
Central infrastructure can make those requirements easier to manage consistently. Instead of every department independently deciding how an AI service should be introduced, the hospital can establish organisation-wide technical and governance conditions while still allowing individual teams to develop applications suited to their work.
Charité CEO Prof. Heyo K. Kroemer describes secure AI infrastructure as an important foundation for the hospital's wider digital transformation and for future AI-supported processes. Following the test phase, access to Clé will now be expanded step by step across the workforce.
Hospitals are becoming AI infrastructure operators
The development in Berlin reflects a broader shift in healthcare AI. The first wave of adoption focused heavily on individual algorithms and whether they could outperform or support humans on specific tasks. The next challenge is increasingly organisational: how can a hospital make multiple AI capabilities available at scale without losing control over data, security and governance?
That requires a different type of investment. Hospitals need infrastructure capable of connecting models, applications and existing information systems, but they also need identity and access management, monitoring, policies and processes for deciding which AI services are appropriate for particular types of information and clinical activity.
European healthcare organisations face an additional layer of complexity as implementation of the AI Act develops alongside existing requirements for medical devices, data protection and the European Health Data Space. A central AI environment does not remove those obligations, but it could provide a more manageable foundation for applying them consistently across a large organisation.
Charité's approach is therefore interesting beyond Germany. Hospitals internationally are deciding whether AI should remain a collection of products purchased by individual departments or become part of the organisation's core digital architecture. Building a shared platform effectively treats AI capability in the same way hospitals increasingly treat cloud, data and interoperability infrastructure: as something that needs to be governed centrally while serving many different applications.
The success of that model will ultimately depend on what happens after employees gain access. The important measures will not be the number of AI models available through Clé, but whether the infrastructure enables applications that improve clinical care, research, education or hospital operations without compromising security and professional responsibility.
Charité is nevertheless taking an important implementation step. Instead of asking where another AI tool can be tested, the hospital is building the infrastructure through which AI can become a managed capability across the organisation.
Add ICT&health on Google
Show more content from ICT&health in Google Search.
