Source: LSE Impact Blog

The LSE Impact Blog features an analysis challenging the assumption that Creative Commons (CC) licenses effectively protect academic work from being used to train generative AI models. The author argues that current copyright frameworks, particularly the text and data mining (TDM) exceptions present in many jurisdictions like the UK and EU, often override the restrictions placed by standard CC licenses. This means that even works licensed as "Non-Commercial" or "No-Derivatives" may still be legally scraped and processed by AI companies under the guise of TDM for research or even commercial development in some contexts.

The post highlights a critical gap between the intentions of researchers who share their work openly and the legal reality of how that data can be exploited by technology firms. It suggests that the "social contract" of Open Access is being strained, as academics see their labor contributing to commercial AI products without their consent or compensation. The article calls for a re-evaluation of licensing structures and potentially new regulatory mechanisms to give authors genuine control over how their works are utilized in the training of large language models.