Copyright and Wrong: The Evolving AI and Copyright Narrative

This session looks at how the global AI and copyright narrative is shaping up, anchoring the discussion in four key developments: (1) the English High Court’s treatment of model weights and “infringing copies” in Getty Images v Stability AI; (2) Germany's opposite approach in GEMA v OpenAI, including its narrow approach to EU text and data mining exceptions and its readiness to treat AI memorisation as legally relevant “embodiment” of works; (3) California’s split approach in Bartz v Anthropic, drawing a hard line between lawful sourcing and pirated acquisition; and (4) new research showing that near-verbatim extraction of in-copyright books from some production models can be feasible even with safeguards, sharpening both infringement and governance risk. 

Against that backdrop, the session explains what Article 53 requires in practice, including a copyright policy and a public summary of training data, and how the Code of Practice pathway is intended to help providers with evidence compliance.  

What you will learn:

• Identify the live fault lines in AI copyright disputes across the EU, the U.K., and the U.S.
• Understand what the AI Act expects from GPAI providers under Article 53 and how the training data summary obligation changes internal data governance and vendor management. 
• Translate recent rulings into a compliance playbook: dataset provenance controls, licensing strategy, and documentary evidence that survives regulatory scrutiny and litigation disclosure.