Renegotiating openness – shaping AI in the public interest

[Pique the interest of your readers with a relevant quote from the document, or SEPTEMBER 2025 15 Katja Mayer, Jochen Knaus, Stefan Skupien et al. Renegotiating openness – shaping AI in the public interest Results of the conference "Yes, we are open?! Shaping artificial intelligence responsibly" Automatic Translation from German with DeepL - preliminary version Renegotiating openness – shaping AI in the public interest ABOUT THE AUTHORS Katja Mayer \\ University of Vienna, Jochen Knaus \\ Weizenbaum Institute, Theresa Züger \\ Humboldt Institute for Internet and Society, Urs A. Fichtner \\ University Medical Centre Freiburg, Katrin Glinka \\ Berlin University of Applied Sciences, Jan Hase \\ Weizenbaum Institute, Lambert Heller \\ TIB - Leibniz Information Centre for Science and Technology, Lucie-Aiméc Kaficc \\ Hugging Face, Sebastian Koth \\ Weizenbaum Institute, Dominik Kowald \\ Know Center Research GmbH, University of Graz, Ilona Lipp \\ University of Leipzig, Katharina Meyer \\ Digital Infrastructure Insights Fund, Petra Ritter \\ Berlin Institute of Health at Charité – Universitätsmedizin Berlin, Anne-Sophie Waag \\ Wikimedia Deutschland e.V., Manfred Hauswirth \\ Fraunhofer FOKUS, Stefan Skupien \\ Berlin University Alliance In collaboration with Lilli Iliev \\ Wikimedia Deutschland e.V., Charlotte Mysegades \\ Weizenbaum Institute Contact:[email protected] ,[email protected] , [email protected] ABOUT THIS PAPER On 27 March 2025 , the one-day conference " Yes, we are open!? Designing artificial intelligence responsibly" took place in Berlin , organised by the Berlin University Alliance (BUA), the Weizenbaum Institute and Wikimedia. Over 100 participants from science, politics, administration, research infrastructures and research management discussed topics related to the tension between open science and AI in two panels and worked on them in a World Café. The assessments and conclusions presented in this position paper are based on the joint work of the authors. Not all co-authors necessarily agree on every point. Likewise, the contents do not necessarily reflect the official positions of the institution(s) to which the authors belong. ABOUT THE WEIZENBAUM INSTITUTE The Weizenbaum Institute is a joint project funded by the Federal Ministry of Education and Research (BMBF) and the State of Berlin. It conducts interdisciplinary basic research on the digital transformation of society and provides evidence-based and value-oriented options for action so that digitalisation can be shaped in a sustainable, self-determined and responsible manner. Renegotiating openness – shaping AI in the public interest Weizenbaum Policy Paper Renegotiating openness – Shaping AI in the public interest Results of the conference " Yes, we are open?! Shaping artificial intelligence responsibly" Abstract Wherever AI is developed, financed and used publicly – in science, administration and public infrastructure – it must be open, transparent, sustainable and oriented towards the common good. This is not a technical detail, but a democratic obligation: public funds require transparency, traceability strengthens trust, and a focus on the common good makes AI a driver of social participation rather than a concentration of private power. For this to succeed, political backing, permanent structures and binding rules are needed. Only clear framework conditions can ensure transparency, access, accountability and participation in the long term. Openness in the age of AI is not optional – it is a democratic mandate. This list of demands is addressed to political decision-makers at federal and state level, in particular to the committees responsible for digitalisation and state modernisation, science and research, economy and energy, and labour and social affairs. It identifies key areas of action in which openness must be rethought and anchored as a guiding principle – so that AI technologies can be put at the service of the public . Renegotiating openness – shaping AI in the public interest Requirements 1 Ensuring binding openness and digital sovereignty 4 2 Embedding sustainability in AI policy 4 3 Financing digital public infrastructure in a sustainable and solidarity-based financing 5 4 Institutionalising cooperation, public governance 5 5 Promote reflective education and research 5 6 Create critical implementation capacities G 7 Official knowledge and clinical AI models as a driver of innovation G 8 Context 7 Renegotiating openness – shaping AI in the public interest \4 For AI design that is oriented towards the common good and responsible, we call on politicians to: 1 Ensure binding openness and digital sovereignty \ Public digital infrastructures must be based on open standards and free open-source software. \ The use of AI in science and the public sector must meet the highest standards of transparency and accountability – e.g. implementation through a public AI transparency register as in the AI Act and AI impact assessments. Open source AI, with openness from training data to software, must be given preference in the public sector, provided that it enables local data processing, traceability and democratic control – including exit strategies from proprietary dependencies. 2 Anchoring sustainability in AI policy \ The development, operation and use of AI systems must be designed to be ecologically, socially and economically sustainable and thus proportionate in terms of achieving objectives and using resources. \ Funding programmes and award criteria for AI systems in research and public procurement must anchor sustainability as a verifiable criterion and operationalise it through appropriate procedures. In this context, government institutions and the scientific community have a role model function for business and society. Renegotiating openness – shaping AI in the public interest \5 3 Digital public infrastructures must be financed on a long-term and solidarity-based financing of digital public infrastructures Infrastructure maintenance: Development, maintenance and data work – whether software, repositories or data standards – are not secondary tasks, but central elements of both digital public services and innovation policy. Therefore: \ Establishment of a solidarity fund for open infrastructures that enables the development, hosting and governance of AI, financed by public funds and contributions from private users and companies. Project-related funding must be supplemented by long-term financing models. 4 Institutionalise cooperation, open up governance \ Establish binding cooperation formats between science, administration, civil society and business (e.g. start-ups). \ Policy-making in the field of digital technologies must be pluralistic, inclusive, genderequitable, accessible, discrimination-sensitive, data protection-compliant, scientifically sound and experience-based. Participatory approaches should be expanded to include the perspectives of those affected in particular. Governance experiences from open science, open source, open platforms and open data sources must be systematically incorporated into AI regulation. 5 Promote reflective education and research Digital and ethical skills must be firmly established in science and administration education – not as an add-on, but as a foundation across all fields and disciplines. Fully funded training courses should be offered at the state level. The opportunity to learn digital skills must be open to all population groups. \ Interdisciplinary courses (e.g. public interest tech, green AI) need targeted funding. Renegotiating openness – shaping AI in the public interest \6 \ Openness, reproducibility, interoperability and sustainability should be incorporated into AI training as central principles; promotion of open science & AI courses. ; promotion of open science & AI courses. There is a need for more space for experimentation at the interfaces between science, civil society and administration. The development and use of AI must be supported by critical, interdisciplinary accompanying research in order to systematically reflect on its impact, risks and transferability and to promote the development of standards oriented towards the common good with the involvement of civil society . 6 Creating critical implementation capacities \ Laws such as the EU AI Act and a national research data law must not only be passed, but also effectively implemented – with a budget, personnel, training and clear responsibilities. The public sector needs its own AI competence centres with interdisciplinary expertise, not least to better understand its own needs and accountability requirements, flanked by the appointment of appropriately trained personnel to strategic administrative positions. \ Principles of open science, such as open source and open data, should be anchored in public administration, education and research in order to make AI transparent, comprehensible and democratically legitimate. 7 Recognise open knowledge and smaller AI models as drivers of innovation Instead of relying on a few large systems, diversified, controllable alternatives are needed and require support for open, participatively developed AI models – especially in the areas of education, law, language and health. \ The responsible use of training data – in compliance with copyright, data protection and ethical standards – and the consistent application of open licences require clear framework conditions and legal certainty. \ This requires bias documentation of the training data used and its continuous improvement through audits that take into account the perspective of those affected . \ Public model hubs are needed, for example for education and justice, in order to provide trustworthy and traceable AI models. Renegotiating openness – shaping AI in the public interest \7 To ensure transparency and quality, open platforms for the evaluation of AI systems must be established. 8 Context Even before the advent of large language models, there has been lively debate in Germany and Europe about the use, openness and orientation of AI and its regulation. Our discussions draw on many sources, including, on the political side, the findings of the German Bundestag's Enquete Commission (2020), the Hiroshima Recommendations for Action (2023), and the OECD Principles (2024), and, on the civil society side, F5 on digitalisation and AI (2024 and 2025), D64 (2024, 2025), AlgorithmWatch (2025), from the field of research, including the Council for Social and Economic Data (2025), funding organisations such as the DFG (2025), but also the position of the German Ethics Council (2023). Artificial intelligence (AI) refers to much more than just chatbots and generative AI, namely data-driven systems that are capable of recognising patterns, supporting decisions or executing actions in a partially automated manner. It is increasingly shaping how we communicate, conduct research, do business and engage in politics. One thing is certain: there would be no artificial intelligence as we know it today – no language models, no image recognition and translation systems, no scientific breakthroughs in the field of machine learning – without open science, free and opensource software, and freely accessible sources of knowledge. Open science and digital commons are the cornerstones of transparent, inclusive and sustainable AI development – they enable democratic control, broad participation and the responsible use of knowledge for all. But this foundation is under threat – from unclear or incomplete legal frameworks, monopolistic market structures and a lack of investment in open, interoperable infrastructures. At the same time, science and civil society have decades of experience: they have built sustainable technical systems, social practices and governance models – from open software projects and collaborative data platforms to global knowledge commons such as Wikipedia. This expertise deserves political recognition – and structural anchoring in the design of public digital infrastructures. Now is the time to make structural openness the guiding principle of digital sovereignty. Renegotiating openness – shaping AI in the public interest \8 Yes, we are open!? Designing artificial intelligence responsibly, 25 March 2025 Event organisation: Katja Mayer, Stefan Skupien, Lilli Illicv, Theresa Züger, Jochen Knaus and Evclin Espcnbcrg Scientific advisory board for the conference: Claudia Müller-Birn (Free University of Berlin), Manfred Hauswirth (Weizenbaum Institute, Fraunhofer FOKUS), Martin Reinhart (Humboldt University of Berlin), Vincc Istvan Madai (Berlin Institute of Health)