A Distributed Approach to Regulating Clinical AI

Share

Researchers from several Harvard University schools and departments and in other international institutions are suggesting the need for an amalgam of centralized and decentralized regulation of clinical artificial intelligence (AI). The researchers discuss the benefits and challenges of such an approach in a report in the open-source journal PLOS Digital Health.

“In the long run,” the report concludes, “we believe that the best approach to regulating clinical AI in practice is to not regulate it centrally, in most instances, but rather to delegate the regulation of clinical AI to local health systems.”

The paper says that while a distributed process involving a robust decentralized process as an adjunct to a centralized regulatory process is optimal, the researchers accept that it is not currently possible. “Such an approach requires the establishment of a specialty of clinical AI and an accountability framework, as well as the development of open data assets, AI registries, and a robust process for public engagement,” they write. “It will also require a shift in regulatory mindset and an acceptance of changes in institutional responsibilities from existing regulatory organizations.”

Read more