Exploring an Innovation Policy for Public AI - Rationales, Examples and Learning

Since the launch of ChatGPT in 2022, AI applications have become even more widespread across personal, business, and public domains. However, AI applications have the potential to extend beyond text generation to improve governance, support decision-makers, and engage in participatory processes. With its broad impact, AI is now seen as a general-purpose technology (Crafts 2021) crucial for societies and democratic processes.
Public AI emphasizes collective involvement in AI’s governance and infrastructure. Key inputs – data, computational resources (compute), algorithms, and human labor – are necessary for meaningful public engagement. Public AI is further defined by three dimensions: trustworthiness (ensuring privacy, fairness, transparency), social innovation (focusing on societal challenges rather than profit), and AI as a common (creating accessible resources and participation). Examples include Estonia’s “bürokratt,” an AI for public services, EU’s GAIA-X, which provides secure cloud services as an alternative to private giants like Amazon, or the non-governmental axolotl AI, which provides an open-source tool for a finetune training of the usability of AI models.
A coherent policy mix is essential to support Public AI, where market and system failures exist and can be resolved by an involvement of the public. Moreover, the use of AI needs an alignment to societal challenges such as sustainability transformations or national technological sovereignty. Governments can offer funding, regulatory frameworks, and educational resources, aligning AI development with societal values and fostering technological sovereignty. Supported by the Mozilla Foundation, Fraunhofer ISI researchers are inviting input from AI practitioners and policy experts to advance a Public AI concept lowering access and use barriers of and fostering investments towards trustworthy and innovative AI applications addressing societal challenges.