EY

The first Global Dialogue on AI Governance needs to clearly establish how this forum will be positioned relative to other international, including UN, initiatives that are already ongoing. The outcomes statement should include a clear future-roadmat for the Global Dialogue, including a leadership/chairing structure that will provide continuity so that decisions and initiatives taken at one meeting carry over and are built upon in subsequent meetings. The chairing of the Global Dialogue could be doen through a combination of a chair and vice-chair, where the chair is rotated out in the subsequent year and replaced by the previous vice-chair, with a new vice-chair onboarded. This would provide a methods for 'handing over the torch' and retaining/building up of institutional memory.

Safe, secure and trustworthy AI;Social, economic, ethical, cultural, linguistic and technical implications of AI;Interoperability of governance approaches;Transparency, accountability, and human oversight;

As AI is being adopted into operational uses in public and private sector with potential for significant consequences in case of failure, it is becoming increasingly vital to have global coordination on quality metrics and quality marks that can signal which AI systems are "safe, secure and trustworthy" for vital tasks. This in turn requires "interoperability of governance approaches" and clear mechanisms for "transparency, accountability and oversight" that is based on an understanding of the "social, economic, cultural and technical implications of AI".

While it may be considered as an integral paft of some of the items that are already listed, a cross-cutting issues is the need for generally accepted metrics and acceptability thresholds in order to avoid arbitrary, or subjective, approaches where the metric/threshold might be post-hoc adjusted to achieve a prior established desired outcome.

The AI Dialogue can most usefully support the advancing of cooperation on AI governance by providing a multi-stakeholder forum for coordination and transparency between existing initiatives. A space for honest exchanges of views where all stakeholders can have an equal voice.

The AI Dialogue should connect with: OECD-GPAI, Hiroshima AI Process, ITU/ISO-IEC/IEEE/CEN-CENELEC, network of AI Safety/Security Institutes, Council of Europe CAI, ASEAN AI working group,

The existing initiative (mentioned in Q14) should provide panels on their ongoing work, to which non-member stakeholders should be invited to provide comments. The emphasis must be on open "dialogue" with all participants (not just the invited panel) from all stakeholder groups.

In addition to the underrepresentation of the Global South and vulnerable communities, the voice of industry actors who are AI adopters (not producers) is generally underrepresented. E.g. when considering reporting on AI risks, industry actors who are adopting third-party AI need to be given a greater voice in being able to clarify the kind of information they need to receive from AI producers in order to be able to make correct judgements about the risks the adopters might be exposing themselves and their clients to, if they choose to adopt a specific AI. The discussions about, e.g. reporting frameworks for AI, are currently dominated by the AI producers who shape the reporting according to their convenience rather than the needs of the adopters.

In order to provide sufficient space for dialogue, at least half he time of each session should be reserved to audience interaction. No sessions that are purely presentations by the panel speakers.