NSF expands AI-ready data funding as GAO flags governance gaps

The National Science Foundation is expanding its effort to make federally supported scientific data easier to find, combine and analyze with artificial intelligence, while an independent audit says the agency still needs clearer rules for reuse and better planning for possible publishing-cost effects.

NSF announced $83 million in Integrated Data Systems and Services awards on July 22, 2026. The investments connect scientific data with computing, instruments, software and AI resources through national-scale projects involving data fabrics, interoperable platforms, secure access and reproducible workflows.

NSF also announced a separate AI Datasets program that could provide between $60 million and $100 million, subject to available funding. The program is aimed at improving existing scientific datasets rather than collecting new data.

The developments create a research-integrity question: Can data governance, provenance, security, reuse rights and public access keep pace with efforts to make more federally supported science usable by automated systems?

NSF is funding the infrastructure first

The $83 million awards are intended to strengthen the systems researchers use to discover, access, share and analyze scientific data. They do not show that the projects have already produced new scientific discoveries, and AI readiness is not a substitute for accurate, complete or scientifically suitable data.

NSF said the awards include projects led by the Morgridge Institute for Research, UC San Diego, UCLA, the University of California, Irvine, the University of Tennessee and the University of Arizona.

The projects include a national data fabric linking repositories and computing resources; platforms designed to support interoperable and reproducible workflows; cloud-based tools for reusable datasets and machine-learning models; and systems that use automated metadata to organize data across disciplines.

In practical terms, NSF is funding shared systems that can help researchers work across datasets and fields instead of requiring each research group to build those connections independently.

A second program could provide up to $100 million

NSF posted solicitation NSF 26-512 on July 21, 2026, and announced the related AI Datasets initiative on July 22. The solicitation anticipates between $60 million and $100 million in awards, subject to the availability of funds. That money has not yet been awarded.

NSF estimates 25 to 50 awards: 10 to 20 impact awards of up to $2 million, five to 10 flagship awards of up to $5 million, and 10 to 20 planning grants of up to $200,000. Full proposals are due by 5 p.m. local time on November 4, 2026.

The program focuses on existing scientific datasets, not new data collection. Eligible work can include metadata generation, feature extraction, integration of multiple datasets, harmonization and automated pipelines for AI-assisted analysis.

Applicants must explain how they will address dataset security and integrity, govern access and contributions, and involve the scientific communities responsible for the data. Those requirements matter because a dataset that is easy for an AI system to process may still contain gaps, inconsistent definitions or limitations that affect the conclusions drawn from it.

GAO identified unresolved public-access questions

The NSF announcements came after a May 21, 2026 report from the Government Accountability Office examined how federal research agencies are implementing a shift toward immediate public access to federally funded publications.

GAO found that NSF’s public-access plan did not fully address use and reuse rights, including how others may share, modify or use NSF-funded publications and what restrictions apply. GAO’s review did not find scientific misconduct, manipulated data or an active security breach. It identified policy and planning gaps.

GAO recommended that NSF ensure its final public-access policy addresses the prerequisites for making publications publicly available by default, use and reuse rights, and restrictions such as attribution. The agency also recommended that NSF analyze how expected publishing-cost increases could affect its research efforts and budget.

GAO described possible cost effects as projections based on historical patterns and assumptions, not a final NSF liability or budget estimate. The report said publication charges have generally risen, but the effects will depend on agency policies, publisher practices and how researchers and institutions pay for open access.

NSF concurred with both recommendations, saying it would identify steps to improve public-access policies and address publishing costs. GAO listed the recommendations as open pending confirmation of agency action.

What researchers and the public should watch

For researchers and universities, the immediate milestone is the November 4 proposal deadline. Applicants will need to show not only how data can be made interoperable and usable by AI, but also how provenance, security, integrity and community governance will be maintained.

For taxpayers and users of scientific information, the next tests are whether NSF selects projects that produce genuinely reusable and reproducible datasets, how the agency defines publication reuse rights, and whether it measures the budget effects of open-access publishing.

The funding announcements establish a significant national direction, but not a completed transformation of scientific research. The practical accountability question is whether the rules and financial planning surrounding the new infrastructure keep pace with NSF’s effort to make more federally supported science available to automated systems.

Sources

Look for updates to this story

Discover more from Interactive News

Subscribe to get the latest posts sent to your email.