ARPA-H Leverages AI to Strengthen Health Data Infrastructure, Security

Advanced Research Projects Agency for Health (ARPA-H) is leveraging AI for key programs — including the agency’s Biomedical Data Fabric (BDF) Toolbox, GRACE chatbot and DIGIHEALS program — to accelerate innovation, bolster data management and advance digital health security, senior agency officials explained at the AFCEA Bethesda HIT Summit last week.

The BDF Toolbox is using AI to enhance biomedical data collection, sharing and analysis. The program builds on National Cancer Institute (NCI) efforts to operationalize data infrastructure platforms. BDF Toolbox sparked workforce interest in generative AI, and since, the agency has expanded to new innovations like agentic AI.

“Agentic AI is no longer a new thing for us. There are so many performers under ARPA-H portfolio who are using [agentic AI] in real world tests,” said Erika Kim, NCI program lead for ARPA-H’s BDF Toolbox. “Some tools [from BDF] have already been requested for deployment in a smaller network of hospitals because they see the value.”

The agency has embraced agentic AI as a key focus of its AI strategy. ARPA-H uses a variety of AI models acting as agents to monitor and check one another, former Director of Data Innovation and Acting CDO Alastair Thomson told GovCIO Media & Research last month.

“How do we use AI to keep it tabs on AI?” Thomson said during a recent GovFocus program. “We work with Microsoft, OpenAI and anthropic, because they’re all coming at these things slightly differently.”

The agency is currently using this approach in their chatbot, GRACE, which functions as an “AI scientist” for program managers and staff. GRACE protects proprietary information while maintaining higher accuracy than outside tools, ARPA-H CIO Nikolaos Ipiotis said during the AFCEA event.

The chatbot can detect hallucinated citations, a problem that has even appeared in recent scientific publications, Thomson said. This is part of a broader effort to use AI to ensure the accuracy and reliability of information, especially in complex scientific fields where a human may find it difficult to keep up with the state of the art, Thomson added. Because users could validate their answers with GRACE, the agency saw increased use of the tool, Ipiotis said.

“The vast majority of our staff is leveraging AI, it can cut down [time by] 20% on document summarization or decrease by 14-fold the time to create a draft document,” said Ipiotis. “It’s enabling that human to do their job a lot faster than normally.”

ARPA-H is also leveraging AI to secure hospital systems. Ipiotis outlined how ARPA-H’s DIGIHEALS program strengthened digital security in health care by funding proposals that address vulnerabilities in medical data and systems.

DIGIHEALS proposals focused on automated medical device patching, ransomware intervention, cognitive health assistants for better data organization, cyber reasoning techniques and electronic health record consolidation. The program also identified limitations of future technology deployments and contributed to the development of new innovations in digital security.

“[DIGIHEALS] is identifying what patches are needed and is applying them automatically to ensure the security of their systems. So there’s a great initiative, especially on the cybersecurity, and we’re looking to not only to use them for our programs, but potentially graduate them to use for internal operations,” said Ipiotis.