Summary
Most government agencies are sitting on data that could improve how they serve residents, but outdated readings of privacy law keep it locked away. In this recap of his Route Fifty commentary, Resultant President John Roach makes the case that regulations like HIPAA and FERPA were written to enable cross-agency data sharing, not block it, and lays out a three-part test for deciding when AI belongs in a government process.
[Estimated read time: 2 minutes]
The data agencies need already exists
Most agencies already have the data they need to serve people better. What’s stopping them usually isn’t technology. It’s a misinterpretation of their own regulations.
That’s the argument Resultant President John Roach makes in his recent Route Fifty article, “Breaking down silos: A practical guide for cross-agency data sharing.” He argues that many of the legal barriers agencies cite aren’t barriers at all. They’re often overly cautious interpretations of regulations that were written to support responsible data sharing.
The regulations were built to enable responsible sharing
“[Laws like HIPAA and FERPA] provide clear guidance on how to share responsibly, yet they’re routinely used to justify inaction,” John states in the article.
He points to guidance from the U.S. Department of Health and Human Services, which describes the HIPAA Privacy Rule as balancing patient privacy with the need to exchange information for treatment and operations. That same principle extends beyond health care. Regulations exist to protect sensitive information while allowing agencies to use data in ways that improve outcomes for the people they serve.
The bigger challenge is organizational, not legal. “General counsels and compliance officers default to ‘no’ when they interpret their primary mandate as keeping agencies out of trouble,” John states. Moving beyond that requires executive leadership to ask a different question: How can we use data responsibly to accomplish our mission?
Start with one problem worth solving
Cross-agency data sharing doesn’t have to begin with an enterprise-wide transformation. John recommends starting with one well-defined use case and a measurable outcome.
A focused project gives agencies an opportunity to establish governance, demonstrate value, and build confidence before expanding into other areas. Instead of debating whether data sharing is too risky, leaders can evaluate what happened, what worked, and where the next opportunity exists.
Three tests for a strong AI use case
The article also offers a practical framework for evaluating AI initiatives. Before introducing AI into a government process, ask three questions:
- Is success clearly defined in terms that matter to the people the agency serves?
- Is this a problem AI is uniquely suited to solve, such as identifying patterns across more information than a person could reasonably analyze?
- If AI gets it wrong, can a person review or correct the outcome before it significantly affects someone’s life?
That third test does a lot of work. It’s the difference between a reasonable place to experiment and a use case that warrants a much higher bar regardless of the upside.
Read the Full Piece
John’s full argument, including how privacy-preserving techniques like federated learning let agencies benefit from shared insights without sharing raw records, is worth reading in full on Route Fifty: Breaking down silos: A practical guide for cross-agency data sharing.
