I would like to share with you our recently published 'Prospective Policy Study on Artificial Intelligence and EU Trade Policy.' I copy you the management summary below:

"Artificial intelligence holds enormous promises for our information civilization if we get the governance of artificial intelligence right. For the EU and the Netherlands within ensuring responsible artificial intelligence is a top priority. With the exception of privacy and personal data protection, the tenets of responsible artificial intelligence is not (yet) codified in EU law. The EU is expected to draw up new rules to provide for ethical and human-centric artificial intelligence.

What makes artificial intelligence even more fascinating is that the technology can be deployed fairly location-independent. Data and machine learning code can be moved across today’s digital ecosystem and the predictive outcomes of an artificial intelligence system can be applied at a distance. The liquidity of artificial intelligence inevitably affects the societies it interacts with.

Cross-border trade in digital services which incorporate applied artificial intelligence into their software architecture is ever increasing. That brings artificial intelligence inside the purview of international trade law, such as the General Agreement on Trade in Services (GATS) and ongoing negotiations at the World Trade Organization (WTO) on trade related aspects of electronic commerce.

The Dutch Ministry of Foreign Affairs commissioned this study to generate knowledge on at the interface between international trade law and European norms and values in the use of artificial intelligence. The study embarked on artificial intelligence with a coherent vision that integrates EU external trade with EU governance of artificial intelligence.

The study makes a number of significant findings:

First, international trade law presumably covers cross-border trade in digital services powered by artificial intelligence. A WTO member’s measure that restricts cross-border digital trade could thus be assessed for its conformity with GATS disciplines. Within the confines of the GATS, a member may adopt measures that are not GATS inconsistent or it may seek to justify a GATS inconsistent measures under one of the exceptions. The study reviews the following measures in a hypothetical challenge under the GATS:

1. Data and/or technology localization;
2. Restrictions of cross-border flows of personal data;
3. Digital security;
4. Technological sovereignty;
5. Mandatory technology transfer requirements; and
6. Other behind-the-border regulations.

Second, EU’s trade policy should better anticipate the challenges in the face of the transnational deployment of artificial intelligence and shouldbe aligned with EU rule-making on artificial intelligence.Aside from the General Data Protection Regulation (GDPR), theEU and member states have not yet exercised their right to regulate responsible artificial intelligence and should guard sufficient space to maneuver under international trade law.

Beginning of 2019, 76 WTO Members announced the launch of WTO negotiations on trade-related aspects of electronic commerce. Without mentioning artificial intelligence, the e-commerce negotiations aim for the multilateralization of new WTO disciplines and commitments relating to e-commerce. New trade rules on e-commerce will also provide for the cross-border supply of artificial intelligence.

The study calls for an open and inclusive deliberation of the interactions between the EU’s e-commerce proposal and EU governance of artificial intelligence. Trade law should not move ahead in setting the rules for cross-border trade in artificial intelligence before the EU could adopt its own rules on artificial intelligence.Future law and policy which has to cope with the liquidity of artificial intelligence systems must identify strategies that interlace European norms and values with cross-border trade of artificial intelligence.

The EU e-commerce proposal notably backs new commitments that protect software source code and restrict a countries’ data and technology localization measures, among others.But should not cross-border digital trade in artificial intelligence be made contingent on a healthy measure of transparency of artificial intelligence systems?EU trade policy should not rule out domestic measures that in the public interest mandate source code transparency, accountability and auditability of artificial intelligence systems.

The study moreover contends that the free data flow commitments in EU’s e-commerce proposal foreclose policy space for state-of-the-art data governance. The free flow of data, which enables cross-border trade in artificial intelligence (upstream), does not necessarily come with reciprocal benefits for countries at the receiving end (downstream). The current discourse lopsidedly emphasizes the free data flows without considering how knowledge and surplus value generated of European data should contribute to public value and societal interests."

With compliments,

Kristina Irion

Institute for Information Law (IViR)

University of Amsterdam