Second revision of draft UNGA resolution for Scientific Panel on AI and Dialogue on AI Governance

Architecture, Building, Office Building, City, Metropolis, Urban, Convention Center, Car, Transportation, Vehicle, Flag, Tower

As negotiations advance among UN General Assembly (UNGA) members in New York, a second revision of a UNGA draft resolution (rev2) has been issued outlining updated proposals for the terms of reference and modalities for the Scientific Panel on AI and Global Dialogue on AI Governance. 

This new draft, published on 4 June 2025, offers fewer details about both the Panel and the Dialogue, compared to previous versions. Some of the main changes are outlined below:

Scientific Panel

  • A clarification is introduced that the summary report produced by the Panel is to be non-prescriptive.
  • Rev2 specifies that the Panel’s two co-chairs must include one from a developed country and one from a developing country, with up to three vice-chairs, emphasising geographic and gender balance. Rev1 called for two co-chairs and three vice-chairs from each of the five regional groups.
  • The appointment process for the Panel is changed in rev2, with the Secretary-General launching a criteria-based open call and recommending a closed list of 40 members to the General Assembly, replacing the Independent Appointment Committee described in rev1.
  • Rev2 modifies the reporting requirements, stating that the Panel will provide an update during an annual interactive dialogue of the General Assembly Plenary, organised by the President of the General Assembly, rather than reporting twice annually to a Plenary on AI as in rev1 (which is removed).
  • Rev2 emphasises the principles of ‘independence, scientific credibility and rigor, multidisciplinary, and inclusive participation’ for the Panel’s work, which were not explicitly stated in rev1.

Dialogue on AI Governance

  • Rev2 no longer includes an explicit reference to the mandate of the Dialogue. Instead, it notes that the Dialogue would discuss international cooperation, share best practices and lessons learnt, and facilitate open, transparent and inclusive discussions on AI governance. It then lists some issues that the discussions would focus on, such as capacit gaps, the development of safe, secure and trustworthy AI systems, and interoperability and compatibility of AI governance approaches.
  • There are no more references to the Dialogue rotating between Geneva and New York. The first dialogue is to be convened in 2026, on the margins of the AI for Good Summit, and the second on the margins of the STI Forum in 2027. Rev1 aimed for a first dialogue in 2025. Instead, rev2 introduces a ‘high-level multi-stakeholder side event’.
  • Rev2 includes a revised provision regarding the secretariat support for both the Panel and the Dialogue. It removes explicit references to specific UN entities (ITU, UNESCO, and the Office for Digital and Emerging Technologies) and instead broadly requests the Secretary-General to facilitate Secretariat support within existing resources and mandate, allowing for more flexible implementation.