U.S. Chamber Files Comments to NTIA on Dual Used Foundation Open Models

Re: Request for Comment, National Telecommunications and Information Administration, Department of Commerce; Dual Use Foundation Artificial Intelligence Models Widely Available Model Weights (89 Fed. Reg. 14,059-14,063, February 26, 2024)

Dear Administrator Davidson:

The U.S. Chamber of Commerce (“Chamber”) appreciates the opportunity to respond to the National Telecommunications and Information Administration (“NTIA”) request for comment (“RFC”) on “Dual Use Foundation Artificial Intelligence Models with Widely Available Model Weights.” We agree with NTIA’s statement that “Artificial Intelligence (AI) has had, and will have, a significant effect on society, the economy, and scientific progress.[1]” Furthermore, we appreciate the RFC’s recognition of how open-source models “could play a key role in fostering growth among less resourced actors, helping to widely share access to AI’s benefits.”[2]

            The Chamber continues to be a strong advocate for using technology to assist small and medium businesses. A Chamber report released last year highlighted that 87%” of small businesses believe that technology platforms have helped their business operate more efficiently[3] and that 71% of them plan to adopt the latest technology, including AI.[4]

We remain concerned with the short comment period provided. Given the strategic importance of AI and its many associated technological and policy complexities, agencies must receive meaningful and robust public input to prevent ill-informed policy developments that could hinder the ability of Americans to reap the benefits of AI and maintain global leadership of this groundbreaking technology. Accordingly, the Chamber, with other associations, corresponded requesting a 60-day extension to allow stakeholders to provide the necessary and thoughtful comments.[5]

Unfortunately, that request for an extension was denied, and therefore, the business community can only provide limited feedback and comments. This denial will not provide the agency the benefit of robust feedback, and the denial will hamper the ability of the agency to assess and weigh the comments received. Accordingly, the agency is in danger of ignoring issues or data raised in the comment process.

I. General Feedback 

Many benefits of open-source technology exist. Open-source technology allows developers to build, create, and innovate in various areas that will drive future economic growth. We already see innovation in marketing, communication, cybersecurity, and medicine, among other fields. Access model weights can be a boon to driving safety and security improvements to artificial intelligence by providing greater transparency, allowing flaws to be quickly identified and patched.[6]

We would further highlight the need for NTIA to make decisions based on sound science and not unsubstantiated concerns that open models pose an increased risk to society.[7]

Trust in technology is fundamental for society to embrace and use technology. For this reason, it is key to educate and prepare the American public through public-private partnerships. NTIA can help prepare the American public for a future in which AI is widely integrated throughout society and ensure that the United States has the necessary skilled workforce to benefit from the technology fully. We also further emphasize that the specific policy questions within the RFC are not just being asked within the United States but globally as well. NTIA must be mindful of the current geopolitical dialogue on this topic.

II. Definitions

The Chamber believes it is critically important for NTIA not to deviate from the definition of the “dual-use foundation model” defined within Executive Order 14110.[8] Potential changes in the definition could impact the scope of NTIA’s work, which could put unnecessary burdens on researchers, academia, and developers who are not developing technology that hits the level of “dual use foundation models with widely available, and the deployers of systems build on these models.” NTIA should also look to harmonize terminology with other industry efforts, such as the G7, AI Alliance, Frontier Model Forum, and academic and non-governmental standard development organizations. A consent lexicon is vital in creating an environment where all AI value chain stakeholders can understand and become familiar with the terminology. 

III. Risk & Risk Management 

The Chamber emphasizes that risk-management processes are still evolving, which is why we support the National Institutes of Standards and Technologies (NIST) work around the Risk Management Framework (RMF) and subsequent companion efforts to develop common standards to address and mitigate potential concerns. Additionally, we note that “risk” is contextual. As indicated in the NIST RMF 1.0, “Risk tolerance and the level of risk acceptable to organizations or society are highly contextual and application and use-case specific.”[9] This is why we believe it is essential for NTIA to focus on the marginal risk, which is context-specific. 

IV. Other Issues: 

The Chamber is concerned with NTIA asking if “risk” should be associated with floating-point operations per second (aka “FLOPs”), as using such criteria does not use the appropriate quantitative metrics based on benchmarks and evaluations.

Finally, it is essential to emphasize that there can be valid reasons for some “artifacts,” such as code, datasets, and model weight, not to be released, as many corresponding factors may make releasing such information challenging. 

V. Conclusion

Although we reiterate our concerns about this RFC’s short comment period, we look forward to working with NTIA and other stakeholders to address these matters and encourage NTIA to look for other opportunities to receive necessary feedback on the current request.

Sincerely,

Michael Richards
Senior Director
Chamber Technology Engagement Center
U.S. Chamber of Commerce


[1] 89 Fed. Reg. 14060 available at https://www.federalregister.gov/documents/2024/02/26/2024-03763/dual-use-foundation-artificial-intelligence-models-with-widely-available-model-weights#print.

[2] Id.

[3] Empowering Small Business: The Impact of Technology on U.S. Small Business at 3 (September 2023) available at  https://www.uschamber.com/assets/documents/The-Impact-of-Technology-on-Small-Business-Report-2023-Edition.pdf.

[4] Id.

[5]Letter to NTIA (March 14, 2014) available at U.S. Chamber Multi-Association Comment Extension Request on NTIA’s Dual Use Open Model RFI | U.S. Chamber of Commerce (uschamber.com).

[6] https://arxiv.org/pdf/2211.14946.pdf

[7] https://crfm.stanford.edu/open-fms/

[8] 88 Fed. Reg. 75194 available at https://www.federalregister.gov/documents/2023/11/01/2023-24283/safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence.

[9] NIST Risk Management Framework at 7 (January 2023) available at https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf