Responsible data sharing for AI: a test bench for EU Data Law

The paper addresses data sharing contracts for AI development and training, entered into by parties participating in AI value chains, in order to assess the incentives for responsible data sharing under European data law. As the governance of AI training data is mainly entrusted to party autonomy and the principles of the GDPR, the analysis first assesses whether contract theory, as it is traditionally conceptualized, is an appropriate tool to promote responsible data sharing. Then the paper delves into the impact of the mandatory data protection framework on data sharing contracts, to conclude that the principle of accountability in its organizational dimension requires the parties to 'contractualize' good data sharing practices. Lastly, the analysis comprises a brief overview of the relevant provisions of the AI Act that deal with data governance and AI value chains, to highlight their impact on the relationship between the entities participating in these value chains.