AI4RA Tip of the Week #22

Sponsor AI policies

With the proliferation of AI across research, it is becoming more important than ever for research administrators to stay up to date on rapidly changing sponsor AI policies. Researchers are finding new ways to incorporate AI into their research, including transforming, summarizing, and extracting text from documents.  RAs often support PIs by double checking sponsor policies to ensure the researchers stay compliant, so today’s tip puts some major sponsors’ policies all in one place for you to review. 

National Science Foundation (NSF)

NSF’s AI policy page serves as the agency’s central hub for its AI strategy and compliance guidance, reflecting a commitment to deploying AI in ways that are ethical, transparent, and aligned with the agency’s core values. The page includes NSF’s publicly posted AI Strategy (required under OMB Memorandum M-25-21), an AI Compliance Plan, and an annually updated inventory of AI use cases across NSF directorates and offices — ranging from proposal similarity analysis and document summarization to reviewer expertise matching and AI-assisted software development. 

For research administrators, the most relevant takeaway is that NSF is actively using AI in its own grant management and review operations, and is required to be transparent about it. The use case inventory is worth reviewing: it gives RAs a window into how NSF is thinking about AI internally, which can inform how institutions anticipate changes to NSF processes. The AI Compliance Plan and Strategy are also useful reference documents when institutions are developing their own AI governance frameworks for NSF-funded work. 

National Institutes of Health (NIH)

NIH’s AI policy page presents a framework of existing policies and regulations that govern the use of AI in NIH-funded research, organized across areas including research participant protections, data management and sharing, health information privacy, peer review, and biosecurity. Notably, NIH does not yet have a single standalone AI policy; instead, the page makes clear that existing regulatory frameworks — including 45 CFR 46, HIPAA, the NIH Data Management and Sharing Policy, and the NIH Genomic Data Sharing Policy — already apply to AI-driven research and must be considered before, during, and after AI use. 

For research administrators, two items on this page demand immediate attention. First, NIH explicitly prohibits peer reviewers from using generative AI tools, including large language models, to analyze or formulate peer review critiques, characterizing it as a breach of confidentiality (NOT-OD-23-149). This applies to any reviewer serving on NIH panels, including faculty at your institution. Second, NIH warns that research data used as input for AI tools could result in unintentional disclosure if that data is sent to an external AI provider, a direct compliance risk for any project involving human subject data, genomic data, or other protected information. Both points underscore why institutional AI infrastructure like MindRouter, which keeps data on-premises, matters for NIH-funded research. 

Other Federal Agencies

Other federal agencies, such as the Bureau of Land Management (BLM), the US Department of Agriculture (USDA), and the Department of Defense (DOD), do not have official AI usage policies meant specifically for researchers or research administrators, though many of them have AI adoption strategies and other internally facing resources. The Department of the Interior (DOI) which oversees the BLM has an official generative AI usage policy (linked below) for responsible use of AI within the department, including use of AI with sensitive data, over public networks, and use in private vs public modes. It also provides usage suggestions for incorporating human oversight and judgement, creating risk management strategies, and being transparent. This serves as an example for how if your specific sponsor doesn’t have an AI usage policy, it is always a good idea to go up to its parent organization if you can. 

Have you seen more? Let us know!

These resources were last verified on September 23rd, 2026. If you know of any other major sponsor policies, or if these ones have become out of date, reply to this thread and share your resources with the rest of the AI4RA community of practice. Through thoughtful collaboration, we can all become better informed. 

How can I learn more about this topic?

AI4RA Office Hours – Occurs every fourth Tuesday at 11 AM PST and includes live Q&A and demos.  

Share your questions or success stories

Post them to the AI4RA community forum. Your tips help the whole research‑admin community thrive.