Technology Business Roundtable Comment Letter to NIST on Developing a Federal AI Standards Engagement Plan

Jun 10, 2019

June 10, 2019

Ms. Elham Tabassi

National Institute of Standards and Technology

100 Bureau Drive, Stop 200

Gaithersburg, MD 20899

Comments of Business Roundtable Submitted to ai_standards@nist.gov

Re: “RFI: Developing a Federal AI Standards Engagement Plan”

Dear Ms. Tabassi, On behalf of the members of Business Roundtable, an association of chief executive officers of leading U.S. companies representing all sectors of the economy, I appreciate the opportunity to respond to NIST’s Request for Information (RFI) on Artificial Intelligence (AI) standards and tools. As developers and users of AI across many different industries and use cases, Business Roundtable member CEOs have a strong interest in the development of a government framework on AI that will spur further innovation and support U.S. AI leadership.

Business Roundtable applauds the emphasis on private sector leadership laid out in the White House Executive Order on Maintaining American Leadership in Artificial Intelligence. Collaboration with the private sector to develop AI technical standards and tools will sustain U.S. AI leadership in the long term. Specifically, Business Roundtable believes that NIST’s framework for AI should reflect three foundational principles:

(1) Public-private collaboration is critical to advancing AI innovation and deployment;

(2) AI applications vary significantly across sectors, and standards and tools should reflect this reality; and

(3) Active U.S. participation in existing international standards setting forums is critical to maintaining U.S. AI leadership.

Within these overarching principles, Business Roundtable recommends that NIST integrate the following considerations into its AI framework.

(1) Public-private collaboration is critical to advancing AI innovation and deployment.   

NIST should work with and leverage existing private sector progress on developing tools for AI. This means building on the current pace of innovation by supporting industry’s development of definitions, best practices and tools for AI. It also means that the U.S. government must take care to avoid premature standards setting for still nascent applications of a rapidly evolving technology. This will help to ensure that standards and tools reflect technical considerations and real-world applications.  

Specific recommendations in support of this principle include: 

  • NIST should coordinate with other government agencies to enable private sector and research community access to diverse and large federal datasets for AI development, as it supports the development and improvement of tools for evaluating and training AI technologies. Federal leadership and coordination here will support high-quality data inputs for improving AI applications and yield shared benefits for AI developers of all sizes across sectors. 
  • NIST should convene diverse private sector, research, and other stakeholders in support of AI explainability and accountability, using its privacy and cybersecurity frameworks as a model. In its convening role, NIST should reference DARPA’s work on explainable AI. NIST should also ensure that the government takes coordinated steps to improve the public’s understanding of increasingly sophisticated systems in the world around them.
  • NIST should articulate federal research and development priorities and objectives related to AI that (1) support and complement ongoing private sector investments, and (2) enable fairness, bias prevention and explainability. 

(2) AI applications vary significantly across sectors, and standards and tools should reflect this reality.  

NIST should develop principles for AI stewardship practices that recognize that trustworthiness and fairness look different across sectors and contexts, where specific needs or requirements may arise based on the nature of the domain. This approach harnesses industry expertise to inform best-fit standards and tools for AI development and avoids overly prescriptive frameworks that could frustrate AI deployment at scale. 

Specific recommendations in support of this principle include: 

  • NIST should develop definitions for key characteristics of trustworthy AI by collecting input from stakeholders at relevant federal agencies, in industry and in research roles. Federal investment in this process will ensure that agencies with the most specific and relevant sector-specific knowledge have appropriate input to the development of definitions that are flexible and applicable to a wide range of use cases. 
  • NIST should promote ongoing private sector efforts to develop context-specific measures for evaluating AI performance. Federal investment in this process will leverage existing progress and build on a body of technically grounded private sector research.  

(3) Active U.S. participation in existing international standards setting forums is critical to maintaining U.S. AI leadership.  

NIST should ensure that the United States is well-coordinated and actively engaged in the development of AI standards and tools on the global stage. This engagement supports strategic and effective positioning with domestic and global partners and helps advance U.S. interests as other countries vie for AI leadership. 

Specific recommendations in support of this principle include: 

  • NIST should promote full and coordinated U.S. participation in NIST should promote full and coordinated U.S. participation in leading international cross-sectoral standards setting bodies (e.g., IEEE, IETF, the ISOC/IEC Joint Technical Committee) and forums (e.g., Partnership on AI, OECD), including by the federal government, private sector and researcher community. Federal investment in these bodies will establish a strong U.S. presence in evolving dialogues on AI standards development on the global stage.  
  • NIST should ensure that full consideration is given to standards resulting from international standards setting bodies and forums. As appropriate, the U.S. government should incorporate international standards for agency AI procurement and use practices. To do so, NIST can use its coordination role to provide U.S. agencies with guidance for adopting AI standards for procurement purposes (e.g., by using OMB Circular A-119).  

Finally, Business Roundtable would like to emphasize the intersection of privacy issues and AI development. To this end, we have called on U.S. policymakers to adopt a national privacy law and developed a detailed framework for legislation. 

Business Roundtable recognizes the promise of AI applications and presents these principles for a federal government framework as CEOs committed to building on U.S. technical leadership.  

Sincerely,  

Denise E. Zheng

Vice President, Technology and Innovation Policy

Business Roundtable