ILTechIG New Developments & New Voices Series: Large Language Models and the Treaty Interpretation Game

Consider this international law puzzle: a spacecraft crash lands on the Moon, surreptitiously carrying tardigrades - microscopic creatures that can survive extreme conditions in stasis. Do these tardigrades constitute “harmful contamination” of the Moon under Article IX of the Outer Space Treaty?

This real-world scenario tested ChatGPT's treaty interpretation abilities with surprising results. When prompted to interpret the treaty as an international lawyer, it produced competent but unremarkable analysis. However, when told it was playing a “game” where principles from the Vienna Convention on the Law of Treaties were “cards” to be played in order to persuade an “audience” of judges, ChatGPT delivered a substantially more persuasive interpretation.

Though this represents just one limited case study, it suggests the need for broader research into how AI tools such as ChatGPT will be used in international legal practice. After all, these tools, trained predominantly on Western, English-language materials, are already shaping how we understand international law. So, will these tools democratize treaty interpretation by making international legal expertise universally accessible? Or will efficiency gains come at the cost of disciplinary stagnation, as backward-looking algorithms perpetuate past conventions rather than enabling progressive legal development?

Speaker: Jack Nelson, Thompson Rivers University

Discussant: Deborah Housen-Couriel, Housen-Couriel Law Offices

Sponsored by the International Law and Technology Interest Group of the American Society of International Law