Fei-Fei Li Senate testimony 2023-09-14 (web)

Citation

Fei-Fei Li, Stanford Institute for Human-Centered Artificial Intelligence (HAI), Stanford University. “Governing AI Through Acquisition and Procurement.” Testimony presented to the U.S. Senate Committee on Homeland Security and Governmental Affairs, September 14, 2023. 5 pages. https://www.hsgac.senate.gov/wp-content/uploads/Testimony-Li-2023-09-14.pdf. Read 2026-10-20.

Bottom line

Li says compute and data needs concentrate frontier AI in “a select few industry players”. Of “32 significant industry breakthroughs in AI last year”, “only three originated from academia, and none from government labs.”
She supports the CREATE AI Act and a National AI Research Resource (NAIRR) to give academic researchers compute and data. This is the federal case for public compute that CalCompute repeats at state level.
She ties public investment to the public workforce: without AI talent, government cannot regulate or buy AI well.

Background

Fei-Fei Li is Co-Director of Stanford HAI. She testified to a U.S. Senate committee on federal AI procurement, “in an individual capacity” (footnote 1).

In 2024 and 2025 she co-led the Governor’s working group on frontier AI (California Report on Frontier AI Policy 2025-06-17). The Wiener SB-53 signing release 2025-09-29 (web) names her as a lead. The Senate Judiciary analysis of SB 53 quotes her on the public sector resource gap (SB-53 Senate Judiciary analysis 2025-04-08).

Key claims

Compute concentration (section V, p. 5).

Because of the vast amounts of compute and data required to train these systems, only a select few industry players can currently work at the frontiers to shape the future of the technology, leaving an imbalance in the AI innovation ecosystem that lacks the diverse voices of academia and government labs. In fact, of the 32 significant industry breakthroughs in AI last year, only three originated from academia, and none from government labs.
Imbalance (p. 5). “there is a deep imbalance in the U.S. AI innovation ecosystem that hinders leadership in the field.”
Public workforce (p. 5).

Without the ability to train AI talent, the federal government will not have the necessary human capital to create meaningful regulation, ensure ethical AI procurement, and be the true AI leader it has the potential to be.
NAIRR (p. 5). “The CREATE AI Act will establish a National AI Research Resource which will ultimately provide the needed computational infrastructure and data resources to allow academic researchers to innovate and train the next generation of AI leaders.”
Moonshot (p. 5). The effort “will have to include infrastructure investments such as the National AI Research Resource, as well as national labs focusing on solving the hardest problems”. It needs “robust collaboration across government, academia, and industry”.
Procurement officials (section IV, p. 4). “The AI Training Act can up-skill procurement officials and equip them with a nuanced understanding of AI capabilities and limitations.” Stanford HAI built a government education program for the U.S. General Services Administration.
Vendor disclosure (p. 4). “Vendors should disclose key information about their systems, including reporting how their AI systems work; how their dataset was designed, collected, and annotated; what potential risks their systems pose”.
Data (section III, p. 3). “we must ensure the availability of high-quality, representative data sets.”

Evidence and method

Written testimony with footnotes. The 32-breakthrough figure comes from the AI Index 2023 report of Stanford HAI (footnotes 11 and 13). The testimony gives no compute figures of its own.

Limits

The testimony is federal. It does not mention California or CalCompute.
It is from September 2023, before the NAIRR Pilot started.
The breakthrough count is second-hand. Cite the AI Index 2023 report for it.
Li speaks as an individual, not for Stanford.

Source file: wiki/sources/Fei-Fei Li Senate testimony 2023-09-14 (web).md, updated 2026-10-20.