Open Source Summit Europe 2022

Publicly available datasets are one of the key drivers for commercial AI software. The use of these datasets are governed by dataset licenses. Dataset licenses outline the rights one is entitled to on a given dataset and the obligations that one must fulfil to enjoy such rights. Failure to comply with the license terms might result in serious legal issues. However, unlike standardized Open Source Software (OSS) licenses, existing dataset licenses do not clearly outline the rights and obligations associated with their usage. Also, datasets can be created from multiple data sources each of which may have different licenses which further compounds the issue. Hence, existing approaches on checking OSS license compliance cannot be used. In this talk, we introduce our project OpenDataology, an open source initiative that proposes a new approach to assessing the potential license compliance violations associated with a dataset. We elaborate how our OpenDataology project can be a crowd-sourced medium that enables identifying and documenting the license compliance risks associated with using publicly available datasets for AI software. We demonstrate how OpenDataology identified potential risks of license violations associated with 6 commonly used publicly available datasets in the wild.