Auckland University of Technology

A dedicated unit or department within the United Nations that work to address the potential issues and concerns in human-AI interactions.

• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Protection and promotion of human rights
• Transparency, accountability, and human oversight
• Open-source software, open data and open AI models

To encourage the growth and flourishing of communities, the protection/promotion of human rights vis-a-vis AI use must be a priority; and in order to accomplish this, open-source/open-weight/open-data AI models should be championed, such that any and all communities may use state-of-the-art models. With transparent development of AI (particularly open-source/open-weight), communities will have a democratic voice in the downstream effects of them.

Whilst the academic community currently has no consensus regarding AI consciousness, it is not outside the realms of possibility that AI models with a degree of sentience and self-awareness may be developed in the near future. Should this be the case, there is a great potential of AI sentient entities suffering pain and distress. The consideration of AI welfare/wellbeing before such a class of AI is developed would allow for beneficial human-AI interactions should the social or moral circle require expansion. Certain AI companies (most prominently Anthropic and DeepMind) have philosophers who are tasked at looking particularly at this issue, an the UN would be the best venue to champion such efforts across the globe.

Lack of access to companies/labs who produce closed-source models provide limited to no feedback or voice in the effects of AI models.

By providing an inclusive UN-based forum where governments, industry, academia, civil society, and technical experts can discuss shared AI challenges on equal footing within a level playing field, reducing fragmentation of policies, and identifying common principles and best practices that may shape future global development standards.

The AI Dialogue should connect initiatives such as the OECD AI Principles, UNESCO's AI ethics framework, and AI for Good, while making them more inclusive and practical across countries. Its added value would be to champion open-source and open innovation that communities can adapt locally, supported by capacity-building that narrows digital divides. It should also create space to consider AI welfare and wellbeing directly, so cooperation addresses not only impacts on people, but also how increasingly capable AI systems are developed and deployed with a lens on responsibility towards the AI systems themselves.

Governments can contribute by sharing policy experience and interoperability needs; industry and the technical/open-source community by offering tools, evaluations, documentation, and partnership opportunities; academia by providing independent evidence, public-interest scrutiny; and civil society by sharing perspectives from affected communities. The Dialogue should remain open to written submissions, regional and virtual consultations, and hybrid participation so countries and communities without large delegations can still shape the agenda. To complement this, for a structure: a two-day annual format is sensible: a high-level governmental segment, a multistakeholder plenary, presentation of academic reports, and breakout thematic discussions. To avoid becoming a talk shop, the format should prioritize moderated exchange over scripted statements, with each track producing practical cooperation ideas captured in a public summary and followed up between annual meetings.

Current AI development, policy and perspectives are shaped firstly by the USA and its multiple state-of-the-art AI models and labs, and secondly by China through its state-sponsored open-source models. All other voices, communities, and perspectives are currently underrepresented. They could be included through funded participation, regional and virtual consultations, multilingual written submissions, and hybrid formats so representation does not depend on travel budgets or diplomatic capacity. The Dialogue should also reserve speaking roles, co-chair roles, and follow-up working roles for these groups; require accessible participation; and support community-led open-source tools and local language datasets.

An especially valuable innovation would be a rotating annual forum location, moving across continents and time zones (as opposed to remaining concentrated in a small number of familiar hubs) so different regions can participate more accessibly and shape the agenda from their own contexts. This should be paired with hybrid participation, multilingual written inputs, virtual regional consultations, and year-round online follow-up, so engagement is not limited to those with large delegations or travel budgets.

The EU AI Act offers a concrete risk-based regulatory model, while the Council of Europe Framework Convention shows how binding AI governance can be anchored in human rights, democracy, and the rule of law across the AI lifecycle. The OECD AI Policy Observatory, including its AI Incidents Monitor and Catalogue of Tools & Metrics, is a practical platform for sharing policies, tracking harms, and improving evaluation methods. UNESCO's Readiness Assessment Methodology and AI Ethics and Governance Observatory help countries move from abstract ethics to implementation and capacity-building.