CJEU hearing suggests a possible end to the AI input-output distinction
The CJEU hearing in C-250/25 Like Company v Google has raised a significant question for AI copyright analysis: whether the familiar distinction between input and output can still do practical legal work. The focus is not on the full range of issues in the case, but on one specific point that may reshape how AI-related copyright arguments are framed.
That distinction has often been treated as a useful way of separating what is fed into an AI system from what it produces. In copyright terms, that separation matters because it allows lawyers to ask different questions about the legal status of training material and the legal status of generated output. If the court moves away from that approach, the analysis becomes less compartmentalised and more focused on the overall use of protected material in the AI process.
For copyright practitioners, the significance lies in how the court characterises the relationship between input and output. A rigid dichotomy can obscure the fact that the two stages are connected, and that legal issues may arise across the full chain rather than at only one end of it. If the court accepts that the distinction is artificial or incomplete, parties may need to reassess how they plead infringement, how they identify the relevant act, and how they present causation and materiality.
The practical implication is that AI copyright disputes may no longer be argued by isolating training data from generated content as though they were unrelated events. Instead, the legal analysis may turn on whether the system’s operation as a whole engages copyright-relevant acts. That would make the boundaries of liability less straightforward and increase the importance of a careful, fact-specific assessment of how material enters, is processed by, and emerges from the system.
For UK lawyers, the immediate point is not that UK law has already adopted this position, but that the CJEU’s approach may influence how similar arguments are developed and resisted in copyright disputes with an AI element. The risk is that relying on the input-output divide as a fixed analytical tool may become less persuasive if courts increasingly treat it as too narrow to reflect the legal reality of AI use.
Disclaimer: This post is for general information only and does not constitute legal advice. Specific advice should be sought for your particular circumstances.
Source: