A formal intervention submission by XRSI to the United Nations Global Dialogue on Artificial Intelligence Governance

This intervention was submitted in response to the stakeholder consultation process of the UN Global Dialogue on Artificial Intelligence Governance, established pursuant to the Global Digital Compact and UN General Assembly resolution of 26 August 2025. Learn more at un.org/global-dialogue-ai-governance.

Artificial intelligence is no longer just a tool. It is becoming an actor. Agentic systems are planning, acting, and coordinating across environments. World models are simulating complex realities. At the same time, AI is shaping geopolitical dynamics, influencing cyber operations, information ecosystems, and decision making in high stakes environments. These systems are entering homes, interacting with children, and influencing cognition, behavior, and identity. Capability is accelerating. Governance is not.

This convergence has created a widening trust vacuum. The world does not have an AI capability gap. It has a trust vacuum. And governance must move from principles to proof.

This document represents XRSI’s formal intervention submission to the United Nations Global Dialogue on Artificial Intelligence Governance, established by the UN General Assembly on 26 August 2025. The first convening takes place in Geneva on 6 and 7 July 2026, followed by a second session in New York in 2027. XRSI was invited to submit inputs and recommendations to inform this process.

Prepared by Kavya Pearlman, this intervention advances a concrete, operational path toward trustworthy AI governance at the global level. It calls for minimum global guardrails for agentic AI and world models, with enforceable requirements for human intervention, decision traceability, and accountability by design. It calls for verifiable transparency over self-attestation, the highest protections for children and vulnerable populations, interoperable governance anchored in human rights, and auditability embedded directly into system design and deployment.

The intervention identifies four critical cross-cutting issues: agentic AI and autonomy governance, world models and simulation-driven decision making, data exposure and inferred data risks, and child and developmental safety in AI-mediated environments.

The core shift is clear: from principles to protocols, from claims to verification, and from fragmented approaches to interoperable systems. Trust must be measurable, enforceable, and continuous. The defining question remains: do we retain meaningful human agency over the systems we create? XRSI’s intervention advances a path to ensure the answer is yes.