Generative AI and copyright in the EU and the USA: a view from international investment law

Generative artificial intelligence has disrupted traditional copyright frameworks by enabling large-scale ingestion of protected works for model training, sparking litigation and regulatory responses in the USA and the EU. While existing debates focus on domestic law doctrines such as fair use and text-and-data-mining exceptions, this article explores an underexamined dimension: the potential role of international investment law in disputes arising from state measures—or omissions—affecting copyright holders.Many international investment agreements explicitly include IP within the definition of ‘investment’, raising the possibility that failures to enforce copyright or permissive regulatory stances toward AI training could trigger claims under standards such as full protection and security, fair and equitable treatment or indirect expropriation.Drawing on arbitral jurisprudence and doctrinal analysis, the article evaluates the jurisdictional threshold for copyright-based investments and examines the viability of investor–state dispute settlement claims in this context. It concludes by finding that there are possible scenarios where certain claims could be successful.