US Policy and Regulatory Alert

OVERVIEW

Artificial intelligence (AI) is at the forefront of political discourse, renewing bipartisan support for the United States’ global technological edge. As Congress and the Department of Defense (DOD) seek to increase US AI capabilities, the National Defense Authorization Act (NDAA)—legislation authorizing policy and funding for US military operations and programs passed annually since 1961—will serve as the primary vehicle establishing national security-related AI initiatives for the coming years. 

Use of the NDAA to advance AI development and deployment is not new. In the fiscal year (FY) 2019, NDAA created the National Security Commission on AI, tasking private and government sector stakeholders to make policy recommendations to the administration and Congress related to federal spending on AI research and development, as well as recommendations to streamline DOD’s acquisition systems to accelerate technological innovation.

The Senate and House of Representatives (House) versions of the FY 2024 NDAA include substantial language to further advance AI policies, develop and integrate mission critical AI tools, and streamline AI procurement. As the House and Senate seek to reconcile their individual bills, K&L Gates will examine each version’s unique approach to tackling AI development and optimization. While there are notable policy trends across the House and Senate, each version contains differing focus areas regarding development and deployment of AI capabilities for national security purposes. Negotiators will reconcile differences for a final, conferenced bill when they return to Washington, D.C., in September.

OVERLAPPING LEGISLATIVE PRIORITIES

The majority of defense AI policy emphasizes robust funding and support for research, development, and innovation. Recently, the US government made concerted efforts to robustly support AI grant and contracting opportunities. Both the House and Senate versions of the FY 2024 NDAA expand on these efforts. 

There are many overlapping areas between the two versions of the NDAA. They are similar in their approach to national security-related AI policy, funding, and program development. Both versions call for modernization and integration of AI within DOD and across services and subagencies. Moreover, DOD and Congress prioritize transparency and oversight.

Both bills call for increased training and education related to workforce development including expanding instruction and effective integration and utilization of AI technology.

Most notably, both versions include directives to work with Indo-Pacific allies and partners on AI research and development to support security-cooperation efforts. The Senate version tasks the Secretaries of Defense and State with facilitating cooperative research and development of defense technology with India, specifically prioritizing AI projects. Similarly, the House bill requires the Secretary of Defense to prioritize AI research and co-development with Indo-Pacific allies and partners, which significantly impacts AI development and cooperation between government and industry.

OMISSIONS

Despite overlapping policy trends, there are significant differences in the areas of watermark competition, logistical optimization, and biases. While the House version focuses on specifics around optimization for logistics and defense capabilities, the Senate version takes a broader approach of integration and development and employs competitive development directives.

Unlike the House version, the Senate version authorizes the Secretary of Defense to carry out a prize competition with the sole intention of developing technologies to detect and watermark the use of generative AI. This Generative AI Detection and Watermark Competition would promote market development of generative AI detection technology and increase defense capabilities.

A key element of the House version, absent from the Senate version, is a call for optimization of aerial refueling and fuel management in contested logistics environments. H.R. 2670, 118th Cong. § 345 (as passed by House, 14 July 2023) requires the Chief Digital and Artificial Intelligence Officer to utilize AI for fuel usage, management, and related costs in order to more effectively predict responses to potential challenges. While the Senate version calls for broad integration of AI, it does not emphasize fuel enhancement or predictive modeling. 

Additionally, the House version states the Army’s robotic combat vehicle software pathways program should employ AI navigation software, which the Senate version does not address. While both chambers are interested in the expansion of AI navigation software, specific language will be necessary to ensure congressional intent.

Notably, while both versions acknowledge the need to preserve US leadership in AI-related military capabilities, the House version specifically directs the Secretary of Defense to use AI to identify potential vulnerabilities that can be exploited by known foreign competitors. The House’s stronger position, as it relates to foreign countries, is largely due to the House Select Committee on the Strategic Competition Between the United States and the Chinese Communist Party’s impact on the NDAA legislative process. While the Senate version names China as a significant funding adversary in AI in their report, there is no such language in their version of the bill. This shows the clear intention of the House to assess the risk of China and ensure that the United States is at the forefront of AI military technology.

Finally, while the House version omits language around AI biases, the Senate version prioritizes the issue, requiring the Secretary of Defense to conduct a review of the 2018 DOD AI Strategy to identify, monitor, and mitigate any AI algorithm biases. AI biases will be a contentious subject moving forward as examples of dataset partiality and algorithmic discrimination occur.

REPORT LANGUAGE

Each committee report includes directives regarding AI integration across the military branches. Such mandates include:

• Maritime domain awareness;
• US-Israeli defense cooperation; 
• Dual-use AI and machine learning for predictive maintenance;
• Medical applications;
• Supply chain management; and
• Missile defense, nuclear, and homeland security applications.

TAKEAWAY

The House and Senate share a common goal of more effective integration and modernization of AI capabilities within DOD infrastructure. Additionally, there is a shared interest in increased AI education and workforce development. Given strong bipartisan support, these relatively noncontroversial items are expected to make the final version of the bill in some form. However, there will be some contention over the bills’ discrepancies. As these differences are resolved, K&L Gates will continuously monitor developments through regular engagement with the House and Senate Armed Services Committees and Member offices and can serve as a resource for advocacy goals.