Cross border transfers of personal and non personal data underpin modern digital economies and the development of artificial intelligence (AI) systems. AI model training often requires large and diverse datasets, which leads to frequent transfers of data across jurisdictions. The European Union (EU) General Data Protection Regulation (GDPR) provides several legal bases for such transfers, including adequacy decisions, standard contractual clauses and derogations based on consent. This paper examines how these mechanisms interact with AI model training and analyses their effectiveness in safeguarding data protection while enabling innovation. The study uses doctrinal analysis of EU legal texts, case law and policy documents up to 2018, supplemented by economic literature on data flows, to evaluate the adequacy, consent and standard clause regimes in the context of AI. The results highlight tensions between data protection and the data hungry nature of AI, the challenges of obtaining meaningful consent, and the limitations of contractual safeguards. The paper concludes with recommendations for policymakers and AI developers to enhance cross border data governance and ensure responsible AI model training.