Paper by Ramya Chandrasekhar: “investigates legal entanglements of re-use, when data and content from the open web is used to train foundation AI models. Based on conversations with AI researchers and practitioners, an online workshop, and legal analysis of a repository of 41 legal disputes relating to copyright and data protection, this report highlights tensions between legal imaginations of data flows and computational processes involved in training foundation models.
To realise the promise of the open web as open for all, this report argues that efforts oriented solely towards techno-legal openness of training datasets are not enough. Techno-legal openness of datasets facilitates easy re-use of data. But, certain well-resourced actors like Big Tech are able to take advantage of data flows on the open web to internet to train proprietary foundation models, while giving little to no value back to either the maintenance of shared informational resources or communities of commoners. At the same time, open licenses no longer accommodate changing community preferences of sharing and re-use of data and content.
In addition to techno-legal openness of training datasets, there is a need for certain limits on the extractive power of well-resourced actors like BigTech combined with increased recognition of community data sovereignty. Alternative licensing frameworks, such as the Nwulite Obodo License, Kaitiakitanga Licenses, the Montreal License, the OpenRAIL Licenses, the Open Data Commons License, and the AI2Impact Licenses hold valuable insights in this regard. While these licensing frameworks impose more obligations on re-users and necessitate more collective thinking on interoperability,they are nonetheless necessary for the creation of healthy digital and data commons, to realise the original promise of the open web as open for all…(More)”.