The following is a guest article by Benji Meltzer, Chief Data Officer at Gradient Health

Reliable data is the foundation for building, testing, and validating the tools that will shape the next century of healthcare. From detecting early-stage disease to predicting complications, these tools depend on real-world data to perform safely. Without enough relevant data, progress stalls, models become brittle, biased, and clinically irrelevant. And the gap between AI potential and actual patient outcomes widens.

The Risks of Sharing and Not Sharing  

For many health systems, the conversation about data sharing stops at the risks: regulatory uncertainty, ethical concerns, and the operational burden of making the data available. Those are valid concerns. But they’re only half the equation.

But there’s also a risk in doing nothing.

When healthcare systems opt out of data sharing, they risk their patient population being left out of the datasets that shape tomorrow’s diagnostic tools. That means their patients are less likely to be represented in medical AI models. It means tools may not generalize to their workflows or patient demographics.

The ethical obligation isn’t just about privacy. It’s also about equity. If we don’t include broad, diverse patient data in AI development, we build systems that serve the few, not the many.

Reward Matters  

Research shows that incentives matter when it comes to data sharing. Misalignment with mission, unclear value, or lack of upside all dampen willingness to participate, even when the risk is low.

Yes, financial rewards like revenue-sharing models help, especially when data sharing supports longer-term sustainability. But the most compelling motivator is impact. Systems that share data help shape innovations that directly benefit their patients. They become part of the development loop. They contribute to tools that reflect their populations, reduce misdiagnosis, and improve care quality across the board.

That’s a real, lasting reward for all involved.

The Operational Reality  

Of course, intention isn’t the only barrier. Infrastructure and resource limitations are real.  

Extracting usable data from fragmented and disconnected legacy systems is hard, particularly given the amount of vendor lock-in that exists in our sector, preventing data from being shared at will. De-identifying it to meet regulatory standards can be even harder, and sharing it in a way that supports research, while avoiding custom integrations with dozens of AI developers, can quickly become a full-time job. This is where the model needs to evolve.

Rather than each health system maintaining one-off agreements with every AI company, a more scalable approach is to work through data intermediaries. These groups specialize in taking fragmented, real-world data and turning it into something usable, namely: structured, de-identified, quality-controlled, research-ready, and ethics-governed datasets.

Companies like Gradient Health and others help reduce the operational burden while ensuring that health systems remain in control of how their data is used. It’s less about outsourcing responsibility and more about enabling participation without creating another IT project for data providers.

What to Look for in a Data Intermediary  

If a health system chooses to work with a data partner, it matters who they trust. A good intermediary should:  

• Have a track record of successful data collaborations with health systems and research groups
• Meet high standards for privacy, de-identification, and data governance
• Offer clear, ethical oversight of research projects
• Understand how to curate datasets that reflect real clinical complexity, not just clean cases
• Provide infrastructure that integrates with, rather than complicates, existing IT systems

The goal is to make data sharing safer, easier, and more aligned with the mission of care delivery.

Closing the Gap  

Medical AI will only be as good as the data it’s built on. And right now, the best data is sitting in hospitals and outpatient imaging centers across the world, underused, undershared, and often overlooked.

We have the technical tools, the regulatory frameworks, and the relationships to make responsible data sharing possible. What’s needed now is leadership: from imaging executives, CIOs, CMIOs, and others who see that shaping the future of healthcare starts with how we treat the data we already have.

Doing nothing is a choice. And it carries risk.

Doing something, carefully, responsibly, and with the right partners, could shape the next generation of care.   

About Benji Meltzer

Benji Meltzer is the Chief Data Officer at Gradient Health, global experts at connecting health systems wanting to share data to responsible AI developers who need it.