(Full article published on The Diplomatic Affairs, 30 September 2026)
On 28 September 2026, on the margins of the 63rd session of the Human Rights Council, I took part in a side event organized by the European Public Law Organization (EPLO) at the Palais des Nations in Geneva, titled “The Role of AI in Advancing Human Rights and Multilateral Diplomacy in a Fractured World.” The Diplomatic Affairs has published a report on the discussion, which I am happy to share here.
The panel, convened by Ambassador George Papadatos, EPLO’s Permanent Observer in Geneva, also included Dr. Maria Stamatopoulou and Mr. Sher Verick from the International Labour Organization. Rather than asking whether AI will replace diplomats, the discussion treated it as a complementary analytical tool, and focused on what must stay in human hands: setting the legal, human rights and security limits, linking AI-assisted negotiation to social dialogue, and distinguishing the tasks a machine can perform from the judgments it cannot legitimately make. Bias, hallucinations, traceability and accountability were recurring concerns.
I approached the question from the angle of technical standards, which are multilateral diplomacy in its most technical form. Standards are a form of governance that operates without law: they do not prohibit or permit, they make certain behaviours possible, default, difficult or impossible, before deployment.
AI will not be neutral in diplomacy, just as standards are not neutral.
Whether AI strengthens or weakens the human rights dimension of diplomacy depends on design choices made long before any negotiation starts. Three points from my remarks:
Governance is moving behind closed doors. Some of the most consequential human rights questions on AI are increasingly settled in procurement contracts between governments and technology companies, a shift I analyzed in Governance by Procurement. Multilateral standards are one of the few places where these choices can still be made openly.
Standard-makers set the assumptions, standard-takers inherit them. The same risk applies to AI tools in diplomacy: delegations relying on tools built elsewhere, on other languages and other assumptions, become AI-takers. I explored this lens in Standard-makers and standard-takers.
Screening questions can be applied to diplomatic tools. The draft ITU-T guideline for addressing human rights in standard development, not yet adopted, proposes 15 screening questions phrased as engineering questions. Several apply directly to AI used in negotiations: who could be affected, whether data collection is limited to what is necessary, whether users can understand and contest the technology, and whether misuse is anticipated. Accountability also needs infrastructure, which is the purpose of the new ITU-T Study Group 17 Focus Group on Trust and Identity for Humans and Agentic AI: every delegated action must be traceable to a human who answers for it.
Judgment, legitimacy and trust remain human. My thanks to Ambassador Papadatos and EPLO for the invitation, and to my fellow panelists for a rich exchange.
Original article: The Diplomatic Affairs, “EPLO is debating: The Role of AI in Advancing Human Rights and Multilateral Diplomacy,” 30 September 2026