Recently, while flying long-haul business class on a major Korean airline, I ran into a dispute over service compensation.
During the flight I suffered what appeared to be insect bites, and what followed was a process of discussing the cabin’s pest control and seat maintenance, the crew’s response, and the airline’s follow-up and compensation. By the airline’s database, I am probably not a bad customer. I have flown the same airline’s long-haul routes for more than ten years, and over the next nine months alone I already have six tickets booked or on standby. On the surface, I am close to a loyal VIP customer.
Yet as the compensation discussions unfolded, the consumer’s position proved not so simple. Even after an unpleasant experience, and even if the response is unsatisfactory, it is not easy for me to switch to another airline right away.
Once you factor in a schedule shuttling between Seoul and New York, direct routes, mileage, cabin class, lounges, familiarity with the service, language and connectivity, the real alternatives are more limited than one might think. The episode left me with a question. Does the airline see me as a good customer, or as a customer who cannot easily leave?
The More Concentrated the Market, the More Concentrated the Data
In the past, market concentration was mainly a question of price and supply. When a particular company grew more powerful, we asked whether ticket prices rose, whether routes were cut, whether service deteriorated. Market concentration in the AI era adds another layer: how well a company knows its consumers, and how it uses that information in compensation, pricing, customer service and service priorities.
Aviation, finance, telecommunications and platform industries all accumulate vast amounts of customer data. They can analyze when consumers buy, which prices they respond to, how often they complain, what level of compensation satisfies them, and even how likely they are to actually move to another company. Combine this with AI and companies can classify consumers with far greater precision. In the good sense, this enables personalized service. In the bad sense, it allows them to calculate even the consumer’s weak position.
This is not to say that any particular airline has actually used AI in this way. The issue is what may happen in the future. Suppose, for example, that a customer has used a particular airline for many years, already has several future flights booked and realistically has few other options, and the company knows all of this. What then?
The company might treat that customer with even greater care. But it would also have an incentive to conclude that “this customer can’t easily leave” and stop at minimal compensation or a perfunctory response. AI can gauge a customer’s loyalty, but it can just as easily gauge a customer’s captivity.
The Market-Concentration Question Raised by Stanford HAI
Market concentration has also been treated as an important agenda item in a number of recent policy discussions at the Stanford Institute for Human-Centered Artificial Intelligence (Stanford HAI). In foundation models and the AI ecosystem in particular, a small number of companies with vast computing resources, data, talent and capital are likely to occupy an advantageous position. That is why Stanford HAI highlights the potential of open foundation models to enhance innovation and transparency and to ease market concentration.
This discussion also holds lessons for traditional industries such as aviation and finance. AI does not operate only within new technology companies; it can further strengthen the power of existing industries that are already highly concentrated.
The same applies when we look at the combination of Korean Air and Asiana Airlines. Competition authorities inevitably look at traditional criteria such as routes, slots, traffic rights, fares and seat capacity. In the AI era, however, the concentration of data must be examined as well. Years of accumulated boarding records, mileage information, customer tiers, complaint-handling histories, compensation patterns and future booking information can all become benchmarks for how consumers are treated.
In the past, the saying “if you don’t like it, choose another company” held true to some degree. In a market with fewer options, however, the phrase rings hollow. And if companies can use AI to predict whether consumers can actually leave, consumer choice may exist only on the surface. This is the point that antitrust and consumer protection must take a fresh look at in the AI era.
The Law Must Look Not Only at Price but at Choice
Going forward, antitrust and consumer protection law must look not only at prices but also at the imbalance of power created by data and algorithms. Do consumers have real alternatives? Is the data a company holds being used in ways that weaken consumers’ bargaining power? If AI is used in customer service, calculating compensation, price discrimination and deciding service priorities, are the criteria explainable? Can consumers challenge automated decisions?
This is also where my attention as a lawyer is drawn. Consumer protection in the AI era is not simply a matter of protecting personal data. We must look at how data interprets the consumer’s position, and how that interpretation affects contract terms, prices, compensation and service responses. Especially in industries where market options are limited, a company’s use of data can make the consumer’s weak position even weaker.
In the AI era, the vulnerable may not be those who know nothing about data, but those who have been analyzed all too well by it. A company can know what that person wants, and also what that person will put up with. Above all, it can know whether that person is unable to leave easily.
Korea, too, must now change the questions it asks about antitrust and consumer protection. We should ask not only how much of the market a particular company holds, but how that company uses consumer data. We should look not only at whether prices have risen, but at whether consumers still have real choice. We should look not only at whether AI was used, but at whether AI is being used to calculate consumers’ weak position.
Market power in the AI era can operate more quietly and more precisely. Consumers may not even know they are being discriminated against, and companies may call it efficiency. That is why law and institutions must see the structure of power hidden behind the convenience of technology. AI knows what consumers want. But the more important problem is this: AI even calculates that you can’t leave.