The launch of ChatGPT in late 2022 made AI a staple tool for business executives and corporate decision-makers. It is “not something businesses can simply ignore,” said a February research note from Thomson Reuters.
It is especially critical for companies with international footprints, as they plan and manage operations in diverse geographies where governments can have diverging policies and strategies. “It’s important to understand what [AI] is and how it can help companies that engage in global trade,” the report noted.
Benefiting fully from AI in global trade will prove more difficult than in other uses, said a white paper on “AI in Trade Facilitation” from the UN Economic Commission for Europe (UNECE). For one, the cross-border dimension adds new complexities and urgency to the use and regulation of AI.
Bigger challenges
The most prominent challenge facing AI, in general, is having sufficient “amounts of structured and labeled data to effectively learn and identify patterns and build accurate predictions,” the UNECE said. Using AI in global trade requires even more data and more sophisticated classification and organization structures, as data from foreign markets can be more diverse and nuanced.
Another is the “lack of skilled staff, [as] AI systems are relatively new and not many professionals have the [necessary] skills,” the UNECE said. That issue is amplified if the AI operates in foreign emerging markets where locals have the necessary social and economic insights, but few technical skills to train AI systems.
A third factor is how much governments rely on AI. The UNECE explained a “lack of government strategy and legal clarity” means state bodies like customs or port authorities still have paper-based bureaucracies and outdated laws that stifle the benefits companies gain by using AI.
More uncertainty
Global trade amplifies the risks of AI. “There is a reason to believe AI systems will make more mistakes as they cross borders,” said Anupam Chander, a professor at Georgetown University Law Center. “First, AI might be designed for different environments, [making it] unsuited in new social, cultural and legal contexts.” That makes “AI transplants … problematic.”
Another risk comes from “immense commercial pressures to claim the first mover advantage — attracting both media and venture capital,” Chander said. That means “AI is being rolled out before it is ready.” The cross-border scope would only accelerate rollouts as local companies must compete with regional or even global players.
Ultimately, that hurts “the quality of AI’s judgments. [making it] hard to assess because firms have incentives to proclaim the effectiveness of their AI, while individual users cannot amass the overall data necessary to evaluate it,” Chander said.
Another major challenge that AI faces is determining “who is responsible for the consequences resulting from misuse of AI systems [and] actions,” the UNECE said. In global trade, that responsibility is split among stakeholders in different countries, regulated by diverging laws. Accordingly, the arguments over responsibility will likely change based on where and how the AI system is applied, the UNECE said.
Regulating AI
In his book “Big Data and Global Trade Law,” Chander said AI lawlessness, in general, means “your phone [can] listen in without permission and push advertising based on what it hears, [the] music app is selling your movements [and] the social network’s algorithm promotes hate speech because it [gets] more engagement.”
Accordingly, regulating AI use is essential, particularly if it sends its findings to a company overseas. “AI [in global trade] changes the nature, scope and scale of foreign decision-making,” Chander said. “We are entering into a world in which your credit, your job prospects, your insurance claim, the news you read and even the dates you go on are determined by faceless computers in a distant land.”
AI regulation across borders is always tricky, as it only exists in the virtual world, making compliance enforcement difficult, Chander said. Nevertheless, legal oversight is relatively straightforward when local companies use AI to better understand their home markets, as the organization and training data fall under the same legal jurisdiction.
The big difficulties arise when harvesting data from foreign jurisdictions, said Chander. On the one hand, the AI doesn’t need to comply with home country laws as it does not collect data from locals. It also is not regulated by the foreign country where it collects data, as the AI only uses digital data in the public domain or those shared by local data collection companies.
Lastly, correctly classifying AI as a product or a service is essential to properly regulate it in international trade agreements. A policy analysis note on “International Trade and Artificial Intelligence” from the International Institute for Sustainable Development (IISD) said, “AI is incorporated into … goods [such as] self-driving cars and AI robotics … as well as services like e-commerce or mobile phone apps.”
Accordingly, the technology is sometimes classified as a product, while at other times it is part of a service. That causes confusion as AI usage would need to comply with different regulations depending on where it is used.
Challenging integration
Chander noted that integrating AI regulations into trade agreements will be difficult on a “textual [and] conceptual level.” The main issue is that existing trade agreements almost never mention AI. Therefore, adding such regulations would require renegotiating nearly all existing contracts.
On a fundamental level, AI is a “method of doing things,” Chander said. Regulating it under international trade agreements changes its underlying concept as “trade agreements focus on what is actually provided rather than the process used to provide it. If trade law does not scrutinize whether a particular decision made by a company is made by an individual or committee, then why should it pay attention to the decision-making process at all?”
Another problem with regulating AI in international trade agreements is that “an importing government may not be able to inquire about the process by which a product is produced, only evaluating its quality as it arrives at the border.”
One AI law
Another critical challenge facing AI regulation, in general, is the technology constantly “reshapes industries and transforms the global economy,” the IISD policy paper said. That means “trade policy must [continually] evolve to keep pace and ensure equitable growth.” Accordingly, regulations need to be flexible and adaptive as “technological innovation [almost always] progresses too quickly for legislation to keep up.”
Those fast-paced changes cause governments to react differently and at different speeds. IISD said that to “prevent [inevitable] international fragmentation,” it is vital to standardize laws related to using AI in global trade.
That should ultimately fall to the World Trade Organization (WTO). “As the only organization with a near-global mandate to regulate trade, [it] seems like the logical place to forge trade agreements to deal with these policy gaps,” IISD’s paper said.
Additionally, giving the WTO that responsibility would align with the organization’s efforts to boost its free trade vision and mandate. The IISD noted that AI “brings unbridled productivity if only trade barriers can be stemmed as the technology spreads from developed economies” to developing ones.
This article first appeared in September’s print edition of Business Monthly.
