KEI response to NTIA request for comments on AI Accountability Policy, expressing concerns over trade agreement provisions which limit the transparency of source code and algorithms

Posted on by James Love

KEI-NTIA-AI-Accountability-RFC

KEI response to NTIA request for comments on AI Accountability Policy, expressing concerns over trade agreement provisions which limit the transparency of source code and algorithms

RE: Notice by the National Telecommunications and Information Administration on 04/13/2023, Document Citation: 88 FR 22433, Page: 22433-22441; Agency/Docket Number: Docket No. 230407-0093
Date: June 12, 2023

Knowledge Ecology International (KEI) is a non profit organization, based in Washington, DC, with an office in Geneva Switzerland. Our work involves research and evaluation of policies and norms both domestically and internationally. We will comment on one topic relating to AI audits and assessments.

The United States Trade Representative (USTR), working with other federal agencies and departments, has advocated for provisions in several plurilateral trade agreements that limit the ability of the U.S. and other governments to require access to source code of software owned by a person of another Party, or to an algorithm expressed by that source code.

We have seen these proposals in Chapter 14, Ecommerce, Article 14.17, Source Code, of the Trans Pacific Partnership (TPP), in the Trade in Services Agreement (TiSA), and in Chapter 19 on Digital Trade, Article 19.16: Source Code, of the Agreement between the United States of America, the United Mexican States, and Canada (USMCA).

KEI has commented extensively on this issue in the past, and discussed our concerns in this 2015 comment:

The USMCA provision reads as follows:

      CHAPTER 19 DIGITAL TRADE
      Article 19.16: Source Code
      1. No Party shall require the transfer of, or access to, a source code of software owned by a person of another Party, or to an algorithm expressed in that source code, as a condition for the import, distribution, sale or use of that software, or of products containing that software, in its territory.

      2. This Article does not preclude a regulatory body or judicial authority of a Party from requiring a person of another Party to preserve and make available the source code of software, or an algorithm expressed in that source code, to the regulatory body for a specific investigation, inspection, examination, enforcement action, or judicial proceeding,6 subject to safeguards against unauthorized disclosure.

      fn6 This disclosure shall not be construed to negatively affect the software source code’s status as a trade secret, if such status is claimed by the trade secret owner.

It is the view of KEI that the USMCA language and similar provisions are far too restrictive, and lack sufficient exceptions, even regarding the use of software licensed under obligations to make software code public, such as the GNU General Public License (GPL), which is used in countless services and products today, or when governments have a clear interest in protecting the public from misleading, harmful or criminal programs that can be hidden in software programs.

The recent interest in the role of algorithms and AI services has identified a number of risks to society, many of which are addressed in this NTIA request for comments on AI Accountability Policy. KEI believes that in some areas and topics, policy makers should require software code and algorithms to be transparent. This should not be prohibited by provisions in trade agreements.

Transparency of computer code and algorithms should be both an option in national legislation, and the subject of global norms, in some areas. The European Commision has created a European Centre for Algorithm transparency. In the U.S. Copyright Office’s recent listening sessions on Copyright and AI, many of the groups representing authors and artists have called for measures to ensure transparency in AI services, as have many groups working on the topics addressed in this NTIA request for comment on AI accountability, and in other diverse areas, such as the growing right to repair movement, or the proposals to use government procurement and other measures to promote open standards and interoperability.

While views on transparency obligations for AI source code or algorithms are varied and often nuanced, it is a mistake enter into trade agreements that broadly prohibit government from mandating transparency in areas where there is a compelling case for a government intervention, either to mitigate against harm, or to promote a welfare enhancing policy that makes services more trusted, useful or affordable.

James Love
Knowledge Ecology International
https:/keionline.org
110 Maryland Avenue, NE, Suite 511, Washington, DC 20002
james.love@keionline.org