Tech experts and lawmakers on Tuesday emphasized the importance of codifying the National Science Foundation’s (NSF) National Artificial Intelligence Research Resource (NAIRR) pilot as a way for the U.S. to maintain its global leadership in artificial intelligence.
The NAIRR serves as a shared national infrastructure to support the AI research community and power responsible AI use. NSF launched the NAIRR pilot in January 2024.
The pilot has generated significant demand from researchers and educators across the nation, and the agency has made over 150 resource awards in areas ranging from AI innovation and cybersecurity to wildfire detection.
“We simply cannot compete against China if we don’t have a resource like the NAIRR,” Julia Stoyanovich, the director of the Center for Responsible AI at New York University (NYU), told members of the House Committee on Science, Space, and Technology during a Tuesday hearing.
The hearing, held by the Subcommittee on Research and Technology, focused on DeepSeek – a Chinese artificial intelligence company that sent shockwaves through the technology industry with its January launch of DeepSeek-R1.
DeepSeek-R1 is an open-source model that rivals leading American models – and at a lower cost. Witnesses at the hearing warned that Congress must take action, such as codifying the NAIRR, in order for the United States to keep up with China in the AI race.
Specifically, Stoyanovich said the United States needs to “foster an open research environment to close the strategic gap,” including robust funding for AI science, public datasets, model development, and compute access.
“The National AI Research Resource is essential here,” she said. “Providing academic institutions with data, compute, and training is necessary for us to compete.”
Stoyanovich, who is also the institute associate professor of computer science and engineering at NYU, said that she is “a big proponent” of the NAIRR and uses it in her own research. She said the resource has enabled her research group “to do things we could not have done” without it, such as assessing large language models like DeepSeek.
“The NSF has been running [the NAIRR] in pilot, and they have been doing an amazing job, considering the limited resources that they have had, but we need to really have access to the NAIRR at full capacity,” she said.
Tim Fist, the director of emerging technology policy at the Institute for Progress, added that the NAIRR should host American open-source models. He said this would subsidize distribution for startups, researchers, and small firms.
Perhaps more importantly, Fist said this could also make American AI models easier to trust and deploy, which is essential to securing a long-term advantage in the AI race.
Rep. Jay Obernolte, R-Calif., chairman of the subcommittee, noted that he recently led the reintroduction of the Creating Resources for Every American To Experiment with AI Act (CREATE AI Act), which would make the NAIRR permanent.
“This will provide researchers and developers across the [nation with the] computational and data resources they need to create competitive American AI systems that embody our values rather than those of the CCP [Chinese Communist Party],” Rep. Obernolte said.
The lawmaker said he “was pretty disappointed” that the bill did not cross the finish line last Congress. Nevertheless, he told the witnesses, “We’ll continue to push on that. Hopefully we can get that across the finish line this year.”