JusProg e.V. uses artificial intelligence and machine learning in various ways. Here, we provide transparent information about our use of AI
We use AI exclusively ‘in-house’ for the optimisation and operation of our services, but not in the direct use of the JusProg youth protection programmes. Neither children and young people nor parents or schools interact directly with the AI at JusProg.
Website rating
We also use artificial intelligence, amongst other things, for the age classification of websites, known as ‘rating’. An in-house application utilises various AI models to automatically determine age ratings and supplementary information regarding domains (FQDN); this is carried out via the Vertex service in the Google Cloud Console. We currently use Claude Haiku, DeepSeek (hosted in the Netherlands) and Gemini, but we are also testing other AI models and using self-hosted open-source AI applications.
The use of AI for website rating takes place without any transfer of personal data from our dedicated servers, to which users of our child protection programmes have no access. The AI models receive only domain names (FQDN), prompts and supplementary information (MD, etc.) via technical interfaces (API).
The age ratings determined by AI support our human rating team (Netagents) in their work and are used as the default age classification for websites that have not yet been age-classified by the rating team using human expertise, or that belong to the large number of domains which have relatively low click counts and therefore cannot be checked by JusProg staff with reasonable effort. At the same time, we have continued to use our in-house developed systems for automated default age classification based on keywords for several years now. In all cases, the following applies: human rating always takes precedence over the machine.
Technical systems
Our applications and technical systems consist of highly complex programmes running on large server clusters in two German data centres; these were created by human developers and are continuously maintained. However, as part of their daily work, the developers also use various AI tools for simpler programming elements and routine technical tasks, as well as for testing, evaluation and system monitoring. No personal data of users or individual browsing logs (which we do not create in the first place) are transmitted to the AI in this process.
Websites, texts, images and videos
The German-language texts on our websites and other information platforms, as well as on LinkedIn and other channels, have largely been created by human editors. Where texts have been created or enhanced using AI, these have been and will continue to be published only after human review. This also applies to images, graphics and videos created with the aid of AI. Our internal AI policy prohibits the use of AI to create fake content that leads users to believe the content is real – however, should this be deemed appropriate in specific exceptional cases, clear labelling is mandatory.
Translations, including those into English, are generally carried out using Deepl’s AI translation service, and in some cases also using Google Translate. The creation of easy language versions of texts is also supported by AI tools.
Support
Our JusProg support team consists of human staff; there are no AI-generated automated replies. However, support staff may use AI for research and to assist in drafting their replies, provided that these are first checked by a human and, where necessary, adapted and corrected before being sent to enquirers. The use of AI tools must take place exclusively in an anonymised manner, without the enquirers’ personal data.
Work of the NGO and the Secretariat
As part of JusProg’s standard office and accounting activities, we also use AI services, for example, for research, creating Excel formulas and presentations, optimising processes and carrying out routine tasks, as well as in other ways customary in modern office operations. This use is carried out in accordance with the provisions of the GDPR, the Artificial Intelligence Act (EU AI Act) and other legal requirements, as set out in our internal TOMs (Technical and Organisational Measures).