AI Priced It Wrong: Antitrust Just Found Its Test Case

By Harvey Specter Jr. — Law, Business & Power Correspondent

---

$5.5 billion. Write that number down. Not because of what Johnson & Johnson agreed to pay — but because of what it took to get there. Fifteen years. Seventy-six thousand plaintiffs. A corporation with more lawyers than most countries have judges, running a bankruptcy strategy that a federal court eventually saw through and rejected. The settlement announced this week is not a story about talcum powder. It is a story about what happens when a company miscalculates the patience of the other side. J&J bet — with shareholder money and fifteen years of human suffering as the stakes — that claimants would run out of resources before it did. They were wrong. The lesson is not unique to pharmaceutical litigation. It applies to every business that thinks its legal budget is a moat.

But the story I want to talk about today is quieter than $5.5 billion. It has a smaller number attached, it involves hotel rooms rather than cancer diagnoses, and it is the one that will matter more to how business gets done in the next decade than anything J&J signed this week.

A federal appeals court in the United States has revived an antitrust lawsuit against operators of major New Jersey casinos. The allegation is specific: that these operators used artificial intelligence software to coordinate hotel room pricing in ways that inflated costs across competing properties simultaneously. The case had been dismissed at the district level. The appeals court disagreed with that dismissal and sent it back. That is not a procedural footnote. That is a signal.

Here is the legal architecture underneath that signal, and why it matters for anyone running a business in Malta or the EU.

Traditional price-fixing law requires proof of an agreement — explicit or implied — between competitors. You need communication, coordination, a meeting of minds. The historic difficulty of algorithmic pricing cases was always this: if two hotels independently buy the same AI software, and that software independently recommends similar prices based on similar market data, where is the agreement? The algorithms didn't conspire. They converged. For years, that distinction was enough to kill antitrust claims before they got traction.

The appeals court's decision to revive this case suggests that "independent convergence through shared software" may no longer be a complete defence. The court is effectively asking: if multiple competitors all delegate their pricing decisions to the same algorithmic tool, and that tool systematically produces outcomes that benefit all of them at the expense of consumers, does the shared delegation become the agreement? That is a genuinely new question. The answer, when it eventually comes, will reshape how competition law applies to AI-assisted decision-making.

In the EU, this conversation is already further along than most people realise. The Digital Markets Act and the evolving enforcement practice of the European Commission have been moving toward exactly this territory. Article 101 of the Treaty on the Functioning of the European Union prohibits concerted practices — not just explicit agreements — between competitors. The European Commission has explicitly stated that algorithmic coordination can constitute a concerted practice even without human communication. Malta, as an EU member state, operates under that framework entirely. The Malta Competition and Consumer Affairs Authority has jurisdiction over conduct that distorts competition locally, and it operates within the broader EU enforcement network.

What does this mean practically? If you are running a business in Malta — a hotel group, a rental platform, a service provider — and you are using a shared revenue management platform or pricing algorithm with your competitors, you may be inside a legal risk zone that did not clearly exist five years ago. The tool is not the defence. The outcome is the question.

I have seen versions of this play out in smaller rooms than a federal appeals court. A client of mine — a mid-sized hospitality operator — was using a third-party yield management system that turned out to be used by three of its direct competitors. No one had discussed prices. No one had picked up the phone. The software had simply been doing what software does: optimising. When a dispute arose over a commercial tender, the pricing history became a liability no one had anticipated. We resolved it before anyone filed anything, which is always the goal. But the exposure was real, and it surprised people who thought they had done everything right.

The J&J settlement and the casino antitrust revival are two ends of the same lesson: the legal risk you didn't calculate is always the most expensive one. J&J calculated everything except the endurance of the plaintiffs. The casino operators apparently calculated everything except the question of what "independent" pricing actually means when the algorithm