Building Strong AI & Data Partnerships for Economic and Social Impact

Contents

Executive Summary

This AI summit panel discussion explores how data and AI collaboratives can drive economic growth and social good through shared governance, ethical frameworks, and inclusive partnership models. The speakers emphasize that successful collaborations require aligned incentives, trust-building governance structures, equitable data access, and intentional inclusion of marginalized communities—not technical solutions alone.

Key Takeaways

Trust, not technology, is the bottleneck: Governance frameworks, clear rules, and trustworthy intermediaries matter more than computational power. Without addressing governance and incentive alignment, data will not be shared, and AI systems will remain siloed and unrepresentative.
Data innovation is severely lagging: While compute and AI models have advanced exponentially, data governance, stewardship, and sharing mechanisms remain stuck in pilot mode. Urgent investment in technical and governance innovation—separating personal data from models, creating new fiduciary structures—is essential to unlock AI's potential for high-stakes domains.
Collaboration beats sovereignty competition: Geopolitical competition drives AI concentration, secrecy, and exclusion. A global, open-source AI stack with shared data governance and local data control is more achievable and equitable than national technological autarky.
Representation is not optional; it is foundational: Failing to intentionally include women, girls, marginalized communities, and non-Western voices in data collection, AI design, and governance will perpetuate and amplify historical inequalities. Women's leadership and local languages must be baked into collaboratives from the start, not added later.
Impact requires moving beyond data to intelligence and action: Sharing and standardizing data is necessary but insufficient. Success means connecting data analysis to government decision-making, policy intent, and resource allocation at scale—turning data into actionable intelligence that drives real-world change in climate, health, food security, and economic opportunity.

Key Topics Covered

Data governance and trust frameworks — How regulatory instruments (EU Data Act, Data Governance Act) and contractual models enable safe data sharing
Data innovation and stewardship — Limited progress in data-sharing models compared to compute innovation; need for new technical and governance approaches
Inclusive AI development — Ensuring AI benefits reach low-resource settings, women, girls, and underrepresented populations
Data as foundational infrastructure — Data quality, standardization, and metadata standards as prerequisites for effective AI systems
Cross-sectoral collaboration — Bridging gaps between government, private sector, civil society, academia, and communities
Policy-linked data intelligence — Connecting data analysis to government decision-making and resource allocation
Cultural heritage and open data — Digitizing and sharing cultural assets under appropriate licensing and benefit-sharing arrangements
Climate-health-water nexus — Collaborative data systems for monitoring and responding to climate impacts
Gender-responsive AI — Embedding women's leadership, representation, and accountability in AI design from inception
Skills, infrastructure, and access — Closing the AI skills gap and expanding computing infrastructure in the Global South
AI sovereignty and openness — Reframing sovereignty as data control + collaboration rather than technological autarky

Key Points & Insights

Data is the overlooked raw material for AI: Martin Tisnney emphasizes that while compute innovation has dominated AI discourse, data governance—the foundation for trustworthy AI systems—remains underdeveloped. Personal data is insufficiently protected and innovatively managed, particularly in applications like precision medicine.
Governance frameworks enable trust and participation: Ambassador Harry Fur argues that instruments like the EU Data Act and Data Governance Act are not bureaucratic burdens but trust-building tools that create legal certainty and make data sharing attractive by reducing perceived risks around privacy, security, and competitive advantage.
The GPTNL model demonstrates alternative data collaboration: The Dutch large language model works via contractual agreements with data holders (libraries, archives, media organizations) who retain control, receive revenue stakes, and maintain transparency—showing that innovation and responsibility can align.
Innovation in data stewardship lags behind compute innovation: Current models (data trusts, data commons) have progressed minimally in a decade, while AI compute has advanced exponentially. Financing technical and governance innovation in data sharing is critical to unlocking AI's potential for sensitive domains like healthcare and genetic research.
Framing collaboration as "enlightened self-interest" works: Stakeholders participate in collaboratives when they see tangible benefits—new business models, improved services, environmental impact, inclusive growth. This reframes data sharing from a cost to an investment in shared prosperity.
Scaling requires bridging data to decision-making: Alexander Oprinko (UNDP) argues that data alone doesn't drive change; it must translate into actionable intelligence for policymakers. Examples include satellite data improving farmer decisions (DRA platform) and multi-ministry data exchange to reduce leakage and redirect fiscal resources to policy priorities.
Gender and social bias must be intentionally addressed from design inception: Christine Arab (UN Women) and Cecilia Celesti emphasize that basing AI systems on existing historical data perpetuates centuries of inequality unless women's leadership is embedded in design, data sourcing, policy, and accountability mechanisms. The "AI school" approach in Asia-Pacific trains stakeholders on gender-responsive AI governance.
Representation and inclusion require investment, partnership, and accountability at scale: Data collaboratives cannot simply extract data from marginalized communities; they must invest in local languages, south-south cooperation, women's voice, and accountability infrastructure to prevent compounding existing discrimination based on gender, location, and socioeconomic status.
Relationship infrastructure precedes technical infrastructure: Agnes Kiraaga's 23-year experience with Countdown 2030 demonstrates that successful data collaboratives depend first on building trust and working relationships between researchers, communities, policymakers, and industry—more so than on technical capacity or data centers.
AI for Good requires seven integrated pathways: Frederick Warner (ITU) identifies: (1) access to sovereign data and safe data sharing; (2) infrastructure and connectivity; (3) skills and education; (4) governance standards and implementation; (5) inclusive AI ecosystems; (6) diffusion and adoption models (India's digital public infrastructure as a playbook); and (7) focus on real societal challenges (climate, health, food security, education).

Notable Quotes or Statements

"Collaboration doesn't happen automatically. It happens when incentives are aligned, when governance is clear, and when participants trust the system. Trust is the foundation of data sharing."
— Ambassador Harry Fur, Netherlands

"The degree to which user data is the raw material for artificial intelligence" is often overlooked; we need governance innovation around data, not just compute innovation.
— Martin Tisnney, Current AI

"Data that cannot connect cannot create value."
— Ambassador Harry Fur

"If AI is the last thing that us humans ever invent, we better make sure that if AI is for good, it's actually for good."
— Frederick Warner, ITU

"You can't assume [responsible, inclusive AI] are natural reflexes for the fast-moving tech and startup industry. The track record so far has been: we build it and we'll fix it afterwards."
— Frederick Warner, ITU

"It's not always about what AI could do, but what AI should do."
— Frederick Warner, ITU

"Sharing data is easy, but doing it ethically is the real challenge."
— Cecilia Celesti, UNESCO

"Relationship infrastructure is what is important. If you can't build relationships that allow data sharing to work closely, it's going to be impossible to have successful data collaboratives."
— Agnes Kiraaga, African Population Health Research Council

"If you base [AI] on existing data, you'll be replicating centuries of inequality… It has to be intentional."
— Christine Arab, UN Women

"This bridge that is required to link data to decision-making to the creation of choices is very important."
— Alexander Oprinko, UNDP Asia-Pacific

Speakers & Organizations Mentioned

Government & Diplomatic

Ambassador Harry Fur — Ambassador at Large and Special Envoy on AI, Kingdom of the Netherlands
Aryen Hildebrandt — Director General, Federal Ministry for Economic Cooperation and Development (BMZ), Germany

United Nations & International Organizations

Frederick Warner — Chief of Strategic Engagement, ITU (United Nations specialized agency for digital technologies); Chief of Strategy and Operations, AI for Good
Alexander Oprinko — Regional Innovation Adviser and Team Leader, UNDP Asia-Pacific
Dr. Huin Kim — Chief of Hydrological and Water Resources Division, World Meteorological Organization (WMO)
Christine Arab — Senior Policy Adviser, UN Women

Civil Society & Research

Gorov Gudwani — CEO and Co-Founder, Civic Data Lab
Martin Tisnney — Founder and Board Chair, Current AI (AI Collaborative)
Anna Tumo Tudor — CEO, Creative Commons
Agnes Kiraaga — Leads Data Science Program, African Population Health Research Council, Nairobi, Kenya
Dr. Cecilia Celesti — Professor and Adviser on AI, Gender, and Rights; Co-Chair, AI Credentials Group, Women for Ethical AI Platform, UNESCO
Dr. Kathleen Victor — Scientific Cooperation Officer, Pasteur Network
Nasubu — KALA Digital Innovation Lab, Africa (on traffic delay)
Mercedes Forgrass — Leads Innovation and Digitalization Workstream, Partnership in Statistics for Development in the 21st Century (Paris21)

Moderators

Nupura — On behalf of Civic Data Lab

Technical Concepts & Resources

EU Data Act — Regulatory framework for data sharing and trustee intermediaries
EU Data Governance Act — Regulatory framework ensuring clear rules and legal certainty in data sharing
GPTNL (Dutch Large Language Model) — Alternative model using contractual agreements with data holders who retain control and receive revenue stakes
Data trusts — Fiduciary structures allowing communities to share data in controlled environments
Data stewardship models — Alternative governance structures for personal and collective data

Platforms & Initiatives

Civic Data Space — Open-source, use-case-led platform for hosting data and AI collaboratives; includes data exchange, metadata standards, AI model registry, and open datasets
DRA Platform — Open platform using satellite data and machine learning to improve farmer decisions on climate-resilient crops
Thrive — Platform enabling citizen feedback on village-level policy planning
AI for Good Summit / AI for Good Initiative — ITU's annual summit and ongoing initiative (launching July 7–10, Geneva); over 50 UN agencies as partners
Countdown 2030 — Population-based research collaborative across 85 countries tracking maternal mortality and health outcomes (23+ years)
Data Science Without Borders Project — AI collaborative at African Population Health Research Council
AI School (UN Women, Asia-Pacific) — Training program on gender-responsive AI governance and design
Fair Forward Initiative — BMZ-supported initiative sharing 55 open datasets and 16 open AI building blocks as digital public goods
Bashini Initiative — Government partnership (India + Indian Institute of Science) creating digital public goods for language AI and voice technology

Standards & Interoperability

Metadata standards — Ensuring data quality and compatibility across systems
Sectoral data standards — Domain-specific standardization for climate, health, water, etc.
Hamburg Declaration on Responsible AI for the SDGs — Endorsed by 50+ partners; provides shared reference point for fair and inclusive AI development

Skills & Infrastructure

Euro HPC — Initiative expanding access to high-performance computing
AI Factory (Netherlands) — Upcoming infrastructure lowering barriers to advanced AI development for SMEs, researchers, and startups
Digital Public Infrastructure (India model) — Scalable, inclusive approach to technology diffusion cited as playbook for Global South

Methodologies

Frugal science — Community-driven, resource-efficient approach to research; includes communities in data validation and model feedback
Satellite imagery analysis — Using geospatial data for climate, disaster response, waste management
Qualitative data collection — Ground-level engagement with communities to understand local contexts and validate quantitative findings

AI/ML Models & Applications

Large Language Models — GPTNL (Dutch LLM) as alternative to web-scraped models; emphasis on contractual, transparent data sourcing
Machine learning for flood forecasting — WMO pilot in four countries improving flood prediction accuracy
AI for voice analysis — Estonian startup using voice to detect blood sugar levels (AI for Good Innovation Factory example)

Gender & Inclusion Frameworks

Women for Ethical AI Platform — UNESCO initiative with AI Credentials Group
Gender-responsive AI — Framework embedding women's leadership, representation, local languages, and accountability from design inception
AI extractivism — Term describing exploitative data practices; to be avoided in ethical data collaboratives

Policy Implications & Recommendations

Move from competition to cooperation: Reframe AI as a global public good rather than a geopolitical arms race; prioritize shared open-source infrastructure and data governance over technological autarky.
Invest urgently in data governance and stewardship innovation: Fund technical innovations (separating personal data from models, new fiduciary structures) and governance innovations (alternative stewardship models, benefit-sharing frameworks) to match the pace of compute and AI advancement.
Build long-term, relationship-based collaboratives: Move beyond pilots to sustained ecosystems with clear incentives, long-term funding tied to measurable societal outcomes, and relationship-first approaches to trust-building.
Embed gender and inclusion from inception: Require women's leadership, representation of marginalized communities, local language support, and accountability mechanisms in all data and AI collaborative design—not as an afterthought.
Connect data to decision-making and resource allocation: Link data analysis to government policy intent, fiscal space, and resource reallocation; measure success by real-world impact, not data volume or model performance alone.
Expand infrastructure and skills access: Support high-performance computing access, data centers, and reskilling programs in low-resource regions; treat data literacy and ethical AI education as public goods.
Standardize and ensure interoperability: Develop and adopt shared metadata, sectoral, and technical standards to enable data from different sources to connect and create value at scale.
Support cultural heritage digitization and open licensing: Fund and facilitate digitization of archives, libraries, and cultural assets; use appropriate licensing (e.g., Creative Commons) to enable research and innovation while respecting intellectual property and indigenous knowledge.

Limitations & Open Questions

Limited discussion of enforcement mechanisms for governance frameworks and ethical safeguards once collaboratives scale
Insufficient detail on how to handle conflicting interests (researcher publication goals vs. community needs vs. IP protection)
Unresolved technical questions on separating personal data from AI models while maintaining utility
Fiscal constraints in Global South acknowledged but solutions remain aspirational (rely on redirecting existing resources via better data intelligence)
Time constraints prevented deeper exploration of implementation challenges and failure modes

Conclusion

This summit panel articulates a vision of data and AI collaboratives as tools for inclusive, equitable economic growth and social impact. Success requires three shifts: (1) from compute-centric to data-governance-centric thinking; (2) from competition to cooperation; and (3) from technical solutions to relationship and accountability infrastructure. The speakers emphasize that governance, trust, inclusion, and decision-linked intelligence matter as much as—if not more than—technological capability. The next phase requires moving from inspiring principles and pilot projects to funded, scaled collaboratives with measurable societal impact and embedded accountability to marginalized communities.