The neuroimaging community is pushing back against a new U.S. policy proposal that would require federally funded researchers to share data only through controlled-access repositories.
Under the policy put forward in December by the U.S. National Institutes of Health, researchers who share human genomic, epigenomic, proteomic or transcriptomic data or imaging data of the face or head must review all data access requests, authenticate the identity of the requesters, use stringent security standards and restrict access for requesters from China, Russia, Iran, North Korea, Cuba and Venezuela, as outlined by the Department of Justice in January 2025. The goal is “to promote maximal responsible human participant data sharing through controlled access while simultaneously responding to emergent privacy and security risks,” according to the NIH’s request for public input on the proposal. The request for comments is open until 18 March.
Neuroscientists are raising multiple concerns about the suggested policy. For one, it is overly broad in scope, according to a written response that the Adolescent Brain and Cognitive Development (ABCD) Study consortium shared with The Transmitter; the group plans to submit the response to the NIH’s request for comment. The type of data that would fall under the policy “would encompass nearly all human subjects research funded by NIH, regardless of whether the data are identifiable or pose any meaningful risk of re-identification,” the group writes.
NIH is requesting comments on a new draft policy that may require human brain imaging data (and other data) from nih-funded research to only be shared via controlled access, & only with certain countries. I suggest neuroimagers read closely & submit comments by 3/18.
osp.od.nih.gov/comment-form…
— Lauren Atlas (@laurenatlas.bsky.social) March 11, 2026 at 9:10 AM
The draft policy also does not line up with the current privacy risks of properly de-identified brain scans, says Anita Jwa, a research scholar in Russell Poldrack’s lab at Stanford University. Jwa is part of OpenNeuro, an open platform for sharing neuroimaging data, and is drafting a response on behalf of that group. “It’s a little bit disproportionate to impose this kind of controlled access to all human neuroimaging data,” she says.
The policy would “be burdensome not just for researchers but also for institutions and data repositories to impose those requirements,” she adds. Currently, most open repositories do not require preapproval and instead grant access once a requester agrees to the repository’s terms of use.
And most repository infrastructure is not equipped to comply with the security standards, which are usually used in contexts involving national security risks, the ABCD consortium writes. “This requirement reflects a fundamental misalignment between the level of risk posed by most research datasets and the level of security required.” The costs of getting repositories up to par with these requirements could reach hundreds of thousands to millions of dollars per institution, the group adds.
The proposed framework would reduce data-sharing, hinder replication studies and disproportionately affect early-career investigators and researchers at under-resourced institutions, the ABCD consortium writes. If implemented, the policy “will result in substantial harm to scientific progress diverting financial resources and time to compliance instead.”
The NIH has not responded to a request for comment.
W
The proposal “seems like a reaction to a problem that doesn’t exist and just serves to restrict things for no reason,” says Mariam Aly, associate professor of psychology at the University of California, Berkeley, and a neuroimaging researcher who is not affiliated with OpenNeuro or the ABCD consortium.
Some neuroimaging datasets contain sensitive information such as questionnaire data, rare-disease indicators or unusual attribute combinations that make participants identifiable, and these may need to be shared under controlled access, Jwa says. But rather than apply the new policy to all neuroimaging and human studies, the NIH should “focus more on providing more specific guidelines, like factors to determine the level of sensitivity of different neuroimaging data.”
Similarly, the ABCD response proposes a risk-tiered framework that distinguishes between identified, identifiable, and properly de-identified data; it would make the security and access requirements proportional to the re-identification risk and align the security standards with research instead of national security risks.
“Any new policy has to be very thoughtful and granular about what kinds of restrictions they impose,” says Tor Wager, professor of neuroscience at Dartmouth College, who uses OpenNeuro and ABCD datasets. “Imposing broad restrictions on all the types of neuroimaging data will have minimal benefit and lots of harm in terms of [scientific] progress.”
Although the changes would not apply retroactively to data already shared, after about a decade of the field moving toward open access and data-sharing, “it’s just a step backward,” Aly says. “I worry about the effects that might have in making people restrict access more than needed—people having to go through all these hoops to access data when they didn’t before, and that just limiting scientific progress that we’ve worked really hard towards.”
