At Ada, we recognise that not all challenges we face can be resolved at national and regional levels alone. Global governance is necessary to ensure that AI technologies and those developing and deploying them are trustworthy, accountable and serve the public.
The UN Global Dialogue on AI Governance – opening in Geneva this July (with a follow-up meeting already scheduled for 2027 in New York City) – is a unique opportunity to discuss shared minimum expectations designed to ensure that the development and deployment of AI uphold the rights, dignity and security of all people.
What makes the Global Dialogue distinctive are three features that are particular to the UN. First, most fora debating AI governance represent a fraction of the world; the UN brings together its 193 Member States and other stakeholders. Second, the UN is the custodian of international human rights law, which makes it uniquely positioned to ground AI governance in existing obligations. Third, the UN carries the mandate for international peace and security, including peacekeeping and humanitarian action: it can speak to AI governance needs across stable and fragile political contexts to preserve and promote the conditions for peace.
The success of the Global Dialogue critically depends on who shows up, the extent to which Member States and other stakeholders engage, and what they ask for: a forum is only as good as the voices that fill it. Civil society and research organisations have both the standing and the responsibility to bring evidence and public interest perspectives into processes like the Global Dialogue. Ada took part in its recent stakeholder consultations and responded to the call for written inputs.
The global AI governance floor
Our submission centres on one organising concept: a global AI governance floor.
The problem the floor solves
AI systems and their supply chains are global and the companies that build them operate across jurisdictions. While governance frameworks that are national and regional stop at borders, AI harms do not.
Fragmentation in AI governance is already visible and at risk of worsening, and regulatory regimes that are hard to reconcile create compliance complexities and opportunity for arbitrage. Regulatory competition for AI investment is real and a race-to-the-bottom dynamic is already emerging. Rights protections that apply only within one jurisdiction are not enough to protect people and communities from harm. Technology crosses borders easily, national rights protections do not. No single national or regional regime can solve this structurally: they need a global response.
What the floor is (and isn’t)
The global AI governance floor is a set of minimum expectations designed to ensure that the development and deployment of AI uphold human rights, advance inclusive and equitable development and economic growth, preserve and promote the conditions for peace, and advance the interoperability of different AI governance regimes.
The floor does not homogenise or fully harmonise AI governance. It is a floor, not a ceiling and gives space to accommodate differences and adjust to contexts above a minimum threshold.
Its legitimacy depends entirely on how it is built – and the Global Dialogue represents an opportunity to build it inclusively, grounded in human rights obligations that already exist and can work as the floor’s foundations.
Three layers
The global AI governance floor is a set of expectations, not a list of principles, and requires three mutually reinforcing components to function.
- Layer one: substantive minimum content indicating what AI must and must not do, what protections must exist everywhere, regardless of jurisdiction, and what capacities and enablers are needed.
The Co-Chairs of the Global Dialogue meeting in Geneva have proposed four thematic clusters to structure conversations around governance; however, how each cluster is interpreted matters. It is important whether, for instance, ‘safe, secure, and trustworthy AI’ is understood narrowly (exclusively long-term, existential or catastrophic risks) or broadly (including nearer-term and already documented harms). Human rights are the connective tissue across the four clusters, not one priority among equals.
- Layer two: institutional infrastructure to develop and maintain a shared, inclusive and routinely collected evidence base on AI opportunities, risks and harms.
Without a shared evidence base, a floor remains speculative and vulnerable to capture. A shared evidence base enables targeted, proportional and adaptive governance solutions. The UN Scientific Panel on AI is a natural home for this, but its mandate must include (or be supplemented by) an institutionally supported and globally inclusive monitoring mechanism to track AI opportunities, risks and harms in a systematic way.
- Layer three: process commitments with regular review, state reporting and civil society participation.
This is the procedural scaffolding that keeps the floor from becoming static and creates realistic pathways for the floor to evolve as may become necessary.
Public deliberation as essential infrastructure
Critically, empirical evidence alone is not sufficient. AI governance decisions shape the lives of people not represented in expert panels or state negotiations. Without structured mechanisms to give public voices procedural weight as part of the governance floor’s institutional infrastructure (layer two), the floor risks losing legitimacy and becoming obsolete.
Our nationally representative survey shows that 91 per cent of the UK public affirm that AI should treat people fairly and 72 per cent that AI regulation would make them feel more comfortable with its deployment. At the same time, 84 per cent of respondents say that they fear government prioritises the interests of large technology companies over those of the public. And this is not a UK anomaly. Whilst data on public attitudes towards AI is scarce, a 2025 global survey by KMPG on trust, attitudes and use of AI showed a public mandate for national and international AI governance, with 70 per cent of respondents believing that regulation is needed. This is reflective of a disconnect between democratic expectations and governance reality.
Public deliberations must feed directly into governance decisions. The Global Dialogue has a concrete opportunity to institutionalise public deliberation from the outset. It is easier to do this now than to fix later.
What we want to see in Geneva and beyond
The success of the Global Dialogue meeting in Geneva hinges on its ability to establish the structures required to shape the AI governance floor across Member States and other stakeholders in the intervening year, leading up to the 2027 Global Dialogue meeting in New York City, where the process can be wrapped up.
Alongside conversations organised according to the Co-Chairs’ proposed thematic clusters, what we need are:
- Dedicated working groups per each thematic cluster, with mandates to drive progress, timelines and deliverables.
- Inclusive leadership, ensuring geographic, gender, and stakeholder diversity, with Global Majority representation in working group chairs and co-chairs.
- Civil society organisations and research institutes as substantive contributors within the working groups, not as outside observers.
The global AI governance floor cannot be built over the course of two Dialogue meetings – the work done between Geneva 2026 and New York City 2027 is as essential as the meetings themselves. The working groups are the bridge and must be mandated and resourced. The outcome document of the 2026 Global Dialogue in Geneva should be explicit about this continuity.
Regional gatherings and online formats can be used to broaden participation beyond those who can travel to Geneva and New York City, and to maintain momentum. Fellowship programmes, with fellows integrated into the activities of the working groups, can be put in place to enable meaningful participation of underrepresented communities, going beyond symbolic inclusion.
This submission is a beginning, not a conclusion. At Ada, the global AI governance floor is a concept and a call to action. We are actively looking for partners and want to hear from civil society organisations, researchers, and Member State representatives working towards a similar ambition.
Please reach out to Friederike Schueuer, our Head of EU and Global AI Governance (email and LinkedIn)