The price is personal: How AI could kill the free market

Second-year LLB student at KCL, Mary-Ellen Dyson, examines how AI-powered pricing is outpacing UK law

Imagine: it’s a wet Friday evening in November. You’ve had to work until past midnight on an important contract, and it’s now too late to travel home by public transport. So, you decide to book an Uber. You check your phone. You realise you forgot to bring your portable charger today. Your phone’s on 12%. Because your battery is low, Uber is (allegedly) going to charge you more. Uber denies this.

This is called “surveillance pricing”, or “personalised pricing”. It’s when corporations use your personal data, such as your age, shopping habits, and even reportedly your device battery, to decide how much to charge you. AI — the capacity to automate intelligent processes, like decision-making — makes it possible at scale, harnessing massive amounts of data to make decisions in real time.

Personalised pricing is different to “dynamic pricing”, in which prices are rapidly updated to reflect demand, like for airline tickets. In dynamic pricing, a crucial principle of the free market still applies: prices reflect population-level supply and demand, as expressed by buyers and sellers. Personalised pricing is far more insidious: it’s all about the individual.

Personalised prices might seem positive at first, but they’re not good news for consumers. If prices can change based on mutable factors, such as whether you’re standing near the shop, then it becomes impossible to shop around to get the best deals. You never know what they’re charging other customers, so you never know if you’re being unfairly treated. This stops consumers from getting the lowest prices.

Moreover, AI pricing models are designed to optimise revenue for vendors. They exist to find your “pain point”, the maximum that you are willing to pay. This enables personalised price gouging. Presently, the term “price gouging” means dramatically increasing prices in times of greatest need, usually large-scale emergencies. But what if AI could find your times of greatest need? A sophisticated AI model could spot that you purchased Calpol, a condolence card, or headache relief tablets, and conclude that you’re under stress, meaning that you won’t notice a few price increases. It could spot that it’s been more than a month since you last bought sanitary towels, and increase the prices for pregnancy tests. In a more everyday sense, it could spot that you’ve put pasta in your basket, and increase the prices for tomato sauce. If AI systems start seeking to extract the maximum amount of money from every single purchase, everything will become much more expensive for everyone.

The big question is: is this legal?

There are three main areas to consider: data protection, pricing, and competition.

Data protection

Why data protection? Because personalised pricing is enabled by the collection of your personal data. In the grocery sector, for example, this is enabled by your loyalty cards (think Clubcard or Nectar). These collect vast swathes of data about you, which grocers use to make more money from you or sell to third parties. Data collection is far more lucrative than the usual business: whereas grocery retail’s gross margins are approximately 20%, the grocers’ data-based business has gross margins estimated to be as high as 80%. To grocers, your loyalty card is more advantageous than your weekly shop.

Under Article 13 GDPR, an individual has the right to be informed of the purposes of data collection. Grocers like Co-op, Waitrose and Sainsbury’s are very clear in their privacy policies: they will collect your personal data, and it will be used to provide you with personalised offers.

Pricing law

An unassuming statutory instrument just might be able to stop personalised pricing – but it isn’t that simple.

The Price Marking Order 2004 stipulates rules that traders must follow when marking prices. Section 7A, added in 2024, is radical. It stipulates that “a trader … must set out … each selling price and each unit price applicable to the product”.

How could personalised pricing survive this? If each consumer only sees their own personalised price, then the trader is failing to fulfil the obligation to show consumers each selling price applicable to the product.

And, if traders seek to circumvent this by saying that the selling price is “subject to conditions”, section 7A covers this too: if “the availability of a particular selling price is subject to the satisfaction of certain conditions, … the trader must prominently display those conditions … in a form which is unambiguous, clearly legible, easily identifiable … and … in close proximity to the product”.

In a world of personalised pricing, how could this requirement be met? There are so many conditions that an AI model could consider when assigning you a price: your device type, your recent purchases, your battery level, your current location, your home address – and all of those are already occurring today. As AI becomes more sophisticated, the possibilities are endless. Moreover, the opacity of AI models means that traders won’t even know what criteria they are using to assign prices. If the traders don’t even know which conditions the AI model is considering, how can they display these conditions unambiguously? This seems to restrict the possibilities for personalised pricing.

The problem is that section 7A is not being fully enforced. Take in-app personalised offers, which are used at several grocers, including Tesco’s and Sainsbury’s. Through such schemes, customers with loyalty cards are presented in their apps with discounted prices, calculated with their personal data. How does this meet the requirements of section 7A? It’s one of the “selling price[s] … applicable to the product”, yet other customers cannot see it. The conditions through which it is calculated are not “prominently display[ed] … in close proximity to the product”. Under a plain reading of section 7A, personalised offers are illegal. Yet this is not being enforced.

To make matters worse, in a recent report, the Competition and Markets Authority (CMA) acknowledged the use of personalised offers, but failed to suggest that these might be breaking the law. In the same report, they commented on the legality of many other pricing practices. Resultantly, the report implies that the CMA believes that personalised offers are legal. This undermines the credibility of a plain reading of section 7A, and damages any attempt to use section 7A to legally prevent widespread personalised pricing. Laws are meaningless unless they are enforced.

Competition law

UK law is unprepared for personalised pricing’s effect on competition. Just one example is the law surrounding the illegal practice, “predatory pricing”. Predatory pricing is where a trader lowers prices to below cost price in order to price competitors out of the market, and recoups their losses by making prices disproportionately high once competitors are eliminated.

Personalised pricing creates three issues pertaining to predatory pricing.

The first is opacity. Personalised, endlessly-changing prices will make it very difficult to uncover ongoing predatory pricing. No consumer will be able to spot predation, because they will only be able to see their own prices. Whilst traders could be legally required to keep a full record of their prices, the volume of data will make this prohibitively difficult. If predatory pricing becomes impossible to identify, then it will become impossible to prevent.

The second is the way that personalised pricing will change the market. Presently, because there are population-level prices, there is only one population-level market. Once personalised pricing means that prices are determined at the individual level, there will be myriad individual-level markets. Legally, predation is understood in reference to its population-level effects on competitors. However, a predatory business will be able to harm competitors by predating within specific individual-level markets without reference to the population-level market. The legal concept of predatory pricing will have to be heavily adapted to fit with the coming commercial reality.

The third is the way that population-level pricing will be eliminated. Some commentary argues that personalised pricing will make the legal concept of predatory pricing unworkable, because current legal definitions assume that all customers are being predated upon simultaneously. This is untrue. Predatory pricing is often targeted at particular customers. However, even if the law is prepared for individual-level prices, if it is unprepared for individual-level markets, then it remains underequipped to address personalised pricing.

Conclusion

Full-scale personalised pricing is only possible because of AI’s computational power. If personalised pricing becomes endemic, it will transform the economy as we know it.
UK law is inadequate to deal with personalised pricing. Data protection law is not protecting our data from enabling it. A provision buried in a statutory instrument on pricing law could help, but it is woefully underenforced. Competition law must be heavily adapted to address the conceptual shifts that personalised pricing causes.
Personalised pricing is coming, and fast. The UK supermarket, Waitrose, has begun rolling out digital price tags in its supermarkets this year. This infrastructure is ideal for personalised pricing – scan a QR code to see your personalised price in the app. Installing digital price tags is expensive; corporations must be expecting a big return.

AI is poised to kill the free market. We’re not ready. But the corporations are.

Mary-Ellen Dyson is in her second year of reading for an LLB in Politics, Philosophy & Law at King’s College London.