{"id":12894,"title":"Who Shapes AI \u2013 and Who Should?","link":"https:\/\/www.reframetech.de\/en\/2025\/05\/20\/who-shapes-ai-and-who-should\/","date":"05\/20\/2025","date_unix":1747721735,"date_modified_unix":1747721735,"date_iso":"2025-05-20T06:15:35+00:00","content":"

Today, the most advanced AI systems are developed and controlled by a small number of private companies – including OpenAI, Anthropic, Google DeepMind, Meta and DeepSeek. These firms not only lead in the development of frontier models, but also control the foundational infrastructure that underpins the AI ecosystem: computing capacity, training data and cloud services.<\/span><\/p>\n

Public AI \u2013 a Public Alternative to Private AI Dominance<\/h2>\n

This concentration of power is not just a technological reality – it is a political challenge. It raises a critical question: who gets to shape the systems that increasingly influence our societies?<\/span><\/p>\n

To address this growing imbalance, we are pleased to introduce our new <\/span>Public AI White Paper<\/strong>.<\/strong> The publication presents a strategic and actionable framework for building an alternative – an approach to AI development and deployment grounded in transparency, democratic governance and open access to critical infrastructure.<\/span><\/p>\n

Private Power, Public Risk<\/b><\/h2>\n

While public debate increasingly centers on the societal impact of artificial intelligence, the capacity to shape its direction remains narrowly held. The absence of robust public or open alternatives means that the most capable AI systems – and the infrastructure that enables them – are governed by a small group of private actors, often without sufficient transparency, accountability, or public oversight.<\/span><\/p>\n

This dynamic is more than a market imbalance. It is a direct threat to the principles of openness, transparency and democratic oversight. When AI systems reflect the incentives and worldviews of a few, their use – and misuse – risks reinforcing structural inequalities and weakening public control.<\/span><\/p>\n

Public AI: A Democratic Countermodel<\/b><\/h2>\n

The <\/span>Public AI White Paper<\/strong> presents a clear alternative. Public AI is not a rejection of private innovation, but a rebalancing of power. It is about ensuring that societies have the institutional and technical capacity to shape AI, rather than merely consume it. It means AI is not just safe for the public, but also shaped by and accountable to it.<\/span><\/p>\n

For Public AI to be viable, it must be grounded in the real constraints of modern AI development. Building competitive systems requires substantial compute resources, access to high-quality datasets and technical talent – all of which are currently concentrated in a small number of firms and jurisdictions.<\/span><\/p>\n

A Vision Grounded in the AI Stack<\/b><\/h2>\n

The white paper conceptualizes AI as a three-layered stack – composed of compute, data, and models – and identifies critical bottlenecks at each layer. These bottlenecks, resulting from corporate consolidation, limit the ability of public actors to pursue independent evaluation, reproducibility, or public-interest deployment.<\/span><\/p>\n

At the core of the proposed strategy is a bold but essential objective: <\/span>ensuring the continued existence of at least one fully open-source AI model with capabilities comparable to proprietary frontier systems<\/strong>. Without such a model, public engagement with cutting-edge AI remains constrained.<\/span><\/p>\n

To achieve this, the white paper outlines three core policy recommendations:<\/span><\/p>\n

    \n
  1. Develop and strengthen fully open-source AI models<\/strong> and the ecosystems around them.<\/span><\/li>\n
  2. Provide publicly accessible compute infrastructure<\/strong> to support the training and use of open models.<\/span><\/li>\n
  3. Invest in talent and capabilities<\/strong> to ensure a sustainable pipeline of public-purpose AI development.<\/span><\/li>\n<\/ol>\n

    Building the Public AI Stack<\/b><\/h2>\n

    The white paper also proposes three policy pathways corresponding to each layer of the AI stack:<\/span><\/p>\n