The AI Act requires providers of General Purpose AI Models to publish a document outlining their training data and practices, with enforcement starting from August 2 2026. Till date (18 September) we have analysed and scored over 70 such documents and discovered issues such as difficulty in discovering documents, missing, unclear or ambiguous information, and in certain cases potentially incorrect information when compared to other documentation. The student will assist with this project regarding the following: (1) Automate discovery and checking for updates of documents e.g. using python scripts; (2) Help analyse documents and quantify problems e.g. a taxonomy of issues we encounter for each field and their prevalence; (3) Help with maintaining the project website https://aial.ie/research/gpai-training-transparency/ and data repository. The project also involves sensitive communications with journalists and regulators to share our findings and discuss next steps. For this, the student will also (4) Help produce ad-hoc reports and blog posts on specific topics.
School of Computer Science and Statistics