AI platforms and large language models (LLMs) enable companies and research organizations to boost efficiency, innovation and competitiveness. However, these services often pose risks to data protection and security. To address this, the Fraunhofer-Gesellschaft—under the leadership of Ingo Weber, Director of Digital Transformation and ICT Infrastructure at Fraunhofer's Munich headquarters—developed FhGenie. This platform has been available to employees since 2023 and is continuously being developed. “Artificial intelligence is a key technology with disruptive potential for innovation. FhGenie demonstrates that Europe has the technological capability to translate this potential into practical benefits. The potential is still enormous, particularly in specific applications for industry, small and medium-sized enterprises and the research sector, as well as with regard to the specifics of AI application use. We must seize this opportunity for Germany as a hub of innovation,” explains Ingo Weber.
FhGenie: Generative AI for everyday work
In June 2023, the Fraunhofer-Gesellschaft introduced FhGenie, a secure ChatGPT-like service that lets employees analyze, create and edit texts while meeting the highest data-protection and information-security standards. A key advantage is that the platform can be customized with organization-specific data to deliver more accurate, context-sensitive results. Today, FhGenie has become an indispensable tool for many Fraunhofer employees. Nearly 12,000 employees use the platform regularly, generating more than a million prompts each month.
Continuous development for security and sovereignty
As a key tool in Fraunhofer's digital transformation, the platform is being continuously developed to meet growing requirements. For security and data protection, the platform operates exclusively within the European Economic Area (EEA), with no data leaving the region. Thanks to its stateless architecture, no information about executed transactions is stored; each request is treated as new. Content filters protect against misuse, and voluntary data contributions for benchmarking are made anonymously. FhGenie also sets standards in data protection and compliance. It operates in line with the GDPR, meaning no user data is stored or analyzed. Clear terms of use and technical safeguards add further security.
In terms of technological autonomy, Fraunhofer's use of its own language models and software ensures sovereignty and gives employees full control over the technologies and systems they use. Regular updates include new language models such as GPT-5, Claude Sonnet 4.5 and Teuken 7B, which was trained from scratch in all of the 24 official EU languages as part of the OpenGPT-X project and developed entirely in Europe. Further updates in recent years include the integration of DALL E 3 for image generation, multi-upload capabilities, automatic file type recognition and ongoing UI improvements.
Advanced features include web and intranet searches, application programming interfaces (APIs) for research projects and on-premises operation on a scalable Kubernetes platform. It also integrates with the SAP Support and Information Portal, an internal knowledge platform, and other internal systems. In addition to these developments, Fraunhofer has built a vibrant internal AI community of more than 4,500 employees who regularly share ideas on AI topics. As an innovation platform, FhGenie also provides an open API for research projects, enabling AI functions to be integrated directly into projects while GDPR-compliant cloud environments such as Microsoft Azure ensure the highest security standards.
Ready for the future: more models, more interfaces
Given rapid advances in AI, plans include expanding the model line-up, for example by adding additional open-source models, to support even more specific applications. At the same time, FhGenie will be integrated more closely into Fraunhofer's existing workflows, including through interfaces to other Fraunhofer systems and external partners. The API functions are also being enhanced to provide targeted support for R&D projects. For example, the imminent introduction of a custom API will enable more flexible integration of FhGenie as well as on-premises data processing. Customizable AI kits are available for specific fields. In addition, features such as load balancing and support for multiple LLMs ensure reliable performance. There is a strong focus on security and compliance. In addition to meeting new regulatory requirements such as the EU AI Act and expanding on-premises variants, an option for handling highly confidential data is under development.
Fraunhofer-Gesellschaft