Attention: You are using an outdated browser, device or you do not have the latest version of JavaScript downloaded and so this website may not work as expected. Please download the latest software or switch device to avoid further issues.
The Core Challenge: Policymakers need data from many sources to answer critical questions, but fundamental barriers prevent effective data sharing. Organizations lose control when they share data copies, and information about the quality of the data gets lost when data is combined.
Start with Questions, Not Data: Too often we answer whatever questions available data can address. Effective policymaking requires identifying the right questions first, then finding ways to access the necessary data sources.
Federal Progress Continues: The National Secure Data Service and National AI Research Resources pilot represent major investments in privacy-preserving data infrastructure that continue moving forward with stable funding, alongside expanded open data availability.
The Bottom Line: Privacy-preserving technologies and federated data architectures offer promising paths for improving evidence-based policymaking while protecting privacy, but realizing this vision requires institutional capacity, leadership commitment, and sustained investment in both technical and governance infrastructure.
In October 2025, I joined Lacey Strahm and Claire Koyle from OpenMined for a panel at the AAAS Science and Technology Policy Forum exploring fundamental barriers to data sharing and emerging approaches to overcome them.
OpenMined is a nonprofit foundation focused on creating open-source technology infrastructure that helps researchers and app builders get answers from data without needing a copy or direct access. Like the Data Foundation, it seeks to expand access to insights from data while ensuring privacy protections for confidential, personal, and proprietary information.
During the panel, we discussed a real and compelling scenario: A policymaker needs to know, "Which communities will face food insecurity due to climate impacts in the next five years?" Answering the question requires integrating weather data, crop yields, transportation infrastructure information, demographics, and economic indicators. Some data is publicly available. Much is not—held by private companies, farmers, state agencies, and organizations with legitimate concerns about sharing.
This pattern repeats across policy domains. Whether addressing chronic absenteeism, water quality, energy assistance, or veterans' health, policymakers need data from multiple sources. Yet fundamental barriers prevent effective integration.
Two core challenges emerged from the discussion:
The Copy Problem: When organizations share data, they lose control over its use. Once transmitted, copies can be redistributed, analyzed in unintended ways, or combined with other datasets in ways that compromise privacy or reveal competitive information.
The Attribution Problem: When data is copied and combined, provenance information is lost. Critical context about collection methodology, quality assurance, limitations, and appropriate uses often fails to accompany the data, making it impossible to assess evidence quality.
OpenMined already mapped more than 3,000 data-holding organizations in the United States that cannot effectively connect their data for policy insights. In the scenario where policymakers look to data to understand which communities are at greatest risk of food insecurity due to climate impacts, siloed data could deprive them of timely, actionable insights about which communities need assistance the soonest or the most. As Lacey Strahm noted, "When we can't connect this data, people suffer."
OpenMined’s Attribution-Based Control (ABC) approach is a framework proposing fundamentally different relationships between data owners and users. The core concept involves distributed architectures where data owners retain their data in place, computational queries come to the data rather than copying it to central locations, and attribution information accompanies outputs.
Such approaches could enable agricultural researchers to analyze farm-level patterns without farmers surrendering proprietary information, or healthcare organizations to conduct multi-state research without exposing patient records. These are promising directions warranting exploration, though questions about scalability, legal frameworks, and practical governance remain.
I emphasized throughout our discussion that data initiatives too often begin with available data and then identify answerable questions. In those cases, data initiatives accept existing infrastructure constraints rather than identify what information is actually needed.
Effective evidence-based policymaking requires the inverse: articulate critical questions first, determine necessary data sources, then address barriers to accessing them. Starting with the questions may seem obvious but runs counter to decades of practice. Emerging privacy-preserving technologies could help support a question-first approach, which can be achieved by moving beyond concepts to tested implementations with clear governance.
Even with the emergence of privacy-preserving technologies, technical solutions alone cannot overcome data access barriers that prevent data initiatives from yielding useful insights for decision-makers. The most persistent challenges are organizational, legal, and cultural. I shared an experience deploying secure multi-party computation in Allegheny County: the technology worked, but much of our time went to working with lawyers, navigating institutional concerns, and establishing governance.
Strong leaders who understand evidence-based policymaking's value can overcome many barriers. The U.S. Commission on Evidence-Based Policymaking exemplified the value of leadership, with recommendations leading to landmark legislation that established governance frameworks, created Chief Data Officer positions, and mandated “open data by default.” The successful and ongoing implementation of those recommendations by the Evidence Commission wasn't primarily about technology but about leadership and institutional commitment.
AI's emergence makes these conversations more urgent and complex. AI systems require training data whose quality directly impacts outputs affecting people's lives. Data quality issues get amplified at scale, provenance matters for accountability, and existing legal frameworks weren't designed for AI's unique challenges.
Privacy-preserving approaches represent one set of tools in a broader toolkit. Evaluating their practical utility, limitations, and appropriate use cases should be part of ongoing policy development.
An audience member asked about practical barriers. The panel agreed:: It's hard, but it's not impossible. Lacey Strahm shared how careful infrastructure design in European social media research avoided triggering General Data Protection Regulation (GDPR) restrictions while providing needed data and protecting privacy appropriately. When we successfully navigate barriers and document precedents, we can also create templates for future efforts.
This incremental approach is essential. Each documented success—including clear explanations of privacy protections, governance, and value generated—makes the next effort more feasible. Coordination across jurisdictions accelerates progress through shared experiences and common frameworks.
In the United States, we are seeing important areas of progress in overcoming barriers to linking data systems and making data open by default. In 2025 alone, there were many noteworthy federal investments that demonstrated marked structural and ecosystem improvements for federal data:
National Secure Data Service (NSDS): Implemented by the National Science Foundation, NSDS provides infrastructure for discovering and linking statistical data across agencies. The Data Concierge concept helps users understand what data exist and how to request access. The service can extend beyond traditional statistical agencies to research-relevant data from EPA, Interior, and others. Funding remains stable with transition to full implementation expected soon.
National AI Research Resources (NAIRR): Building computational infrastructure and secure environments for AI research using sensitive data, the partnership between NSDS and NAIRR combines statistical data infrastructure with AI computational capabilities.
OPEN Government Data Act: Data.gov now references 50,000 more federal datasets than in 2017. Federal agencies are implementing DCAT-3 metadata standards, with full compliance required by September 2026. This "data about data" infrastructure is essential for making federated architectures functional.
Genesis Mission Executive Order: The executive order establishes the "American Science and Security Platform" at the Department of Energy to integrate federal scientific datasets for AI-accelerated discovery, with a 120-day mandate to address data digitization, standardization, metadata, and provenance tracking. It explicitly involves the federal CDO Council in coordinating data integration across agencies, national labs, academia, and private partners to support AI foundation models targeting 20+ national science challenges.
Federal data is public infrastructure powering economic activity, public health, scientific discovery, accountability, and democracy. But infrastructure isn't just what we build—it's about who can access it and how we govern it.
There is reason for optimism. Federal infrastructure investments continue. New approaches are being tested. Communities of practice are forming. Metadata standards are improving. Leadership committed to evidence-based governance exists across programs and different types of decision-makers.
For communities facing climate impacts, health crises, economic displacement, or educational challenges, or other challenges, progress in managing data to be used for improving outcomes matters profoundly.
Nick Hart, Ph.D., is President and CEO of the Data Foundation. This article reflects themes from a panel discussion at the 2025 AAAS Science and Technology Policy Forum held on October 24, 2025, alongside Lacey Strahm (Head of Policy, OpenMined) and Claire Koyle (Associate for Partnerships, OpenMined).
