Unsanctioned manufacturing disguised under the guise of Research and Development
The practice of federal agencies labeling internally developed AI tools as "research and development" (R&D) to bypass procurement requirements can have significant consequences and implications for federal agencies and broader government operations.
Reduced procurement oversight and accountability
Declaring AI tools as R&D may allow agencies to avoid the usual competitive bidding, vendor evaluation, and compliance checks that procurement procedures require. This could lead to less transparency, weakened vendor competition, and a higher risk of acquiring poorly tested or non-compliant AI technologies.
Potential regulatory and ethical gaps
Bypassing procurement rules might lead to insufficient evaluation of AI risks, including security, bias, or operational impacts. Given federal AI strategies emphasize transparency, bias mitigation, and security standards, circumventing these protocols could undermine these goals.
Acceleration of innovation within federal agencies
On the positive side, treating AI development as internal R&D can speed up experimentation and customization of AI tools tailored to agency needs without being slowed by procurement bureaucracy, supporting the federal push to advance AI adoption rapidly and efficiently.
Challenges for standardization and interoperability
If each agency independently develops AI under R&D without coordination, it may create fragmented, non-interoperable systems contrary to federal objectives of shared AI infrastructure and standards, such as those promoted by initiatives like USAi (U.S. AI platform).
Risk of misuse of R&D designation
Agencies might overuse or misuse the R&D classification to circumvent necessary procurement oversight, potentially undermining federal procurement integrity and public trust.
The federal government’s recent AI Action Plan promotes accelerated AI innovation with deregulatory approaches to reduce "bureaucratic smothering," encouraging open-source AI, streamlined processes, and agency-specific experimentation environments (e.g., USAi). However, it also stresses maintaining trust, security, bias mitigation, and national security evaluations. Agencies balancing innovation and compliance must carefully manage the use of R&D labels to avoid undermining these priorities.
The misuse of R&D labels is a threat to the effectiveness, efficiency, and accountability of federal AI programs. The Federal Acquisition Regulation (FAR) Part 12 requires agencies to prioritize Commercial Off-the-Shelf (COTS) solutions over custom builds. Agencies should be required to revalidate their R&D designations at regular intervals, ensuring that tools labeled as experimental are truly being used for experimentation and not scaled like production systems. The government must be a collaborator, not a competitor, and prioritize solutions that deliver real value to taxpayers and warfighters alike.
[1] https://www.whitehouse.gov/articles/executive-order-maintaining-american-leadership-in-artificial-intelligence/ [2] https://www.whitehouse.gov/articles/executive-order-on-the-establishment-of-the-presidential-committee-of-advisors-on-science-and-technology-pcast/ [3] https://www.whitehouse.gov/articles/executive-order-on-maintaining-american-leadership-in-artificial-intelligence-technology-and-services/ [4] https://www.whitehouse.gov/articles/executive-order-on-the-development-of-a-sustainable-artificial-intelligence-framework-for-american-democracy/ [5] https://www.whitehouse.gov/articles/executive-order-on-the-national-security-memorandum-on-the-establishment-of-the-national-security-council-committee-on-artificial-intelligence/
- The misuse of the Research and Development (R&D) label in the federal workforce could undermine priorities such as trust, security, and national security evaluations, as outlined in the federal government's AI Action Plan.
- Bypassing procurement rules in the federal workforce through labeling AI tools as R&D might lead to a lack of vendor competition, with potential consequences that could impact the general news and politics.
- The acceleration of technology innovation within the federal workforce through treating AI development as internal R&D could lead to challenges in standardization and interoperability among different federal agencies, potentially impacting policy-and-legislation and overall government operations.
