Given AI is trained on the work that the public has produced and legally owns and has made available on the internet, should all of these models be nationalised and taken into public ownership too?

A provocative debate initiated by philosopher and author Janne Teller is gaining traction online, questioning the fundamental ownership structure of today's most powerful AI systems. Teller argues that since models like OpenAI's GPT-4, Anthropic's Claude, and Meta's Llama are trained on vast amounts of copyrighted text, code, and images legally published by the public on the internet, there is a strong case for these models to be nationalized and brought into public ownership. This position frames AI not merely as a corporate product but as a derivative work of collective human intellectual output, raising profound questions about equity and access in the age of artificial intelligence.

The argument hinges on established intellectual property (IP) principles, suggesting that the current private exploitation of models trained on public data creates a significant asymmetry. While companies like OpenAI and Google derive immense commercial value, the public—whose creative and intellectual labor forms the training corpus—has no ownership stake or guaranteed access to the resulting technology. This debate intersects with ongoing lawsuits over copyright infringement and the 'fair use' doctrine, pushing the conversation beyond legal compliance toward a broader socio-economic model. If taken seriously, it could fuel policy discussions around public AI utilities, open-source mandates, or new governance frameworks for foundational models seen as critical public infrastructure.