Standing

Not every dispute that reaches a court gets examined on its merits. First, the court clarifies whether the person filing the lawsuit is even allowed to sue. That is exactly what standing means. As a rule, you only have it if you yourself have suffered harm or are directly threatened. Someone who merely believes a company is doing something wrong does not yet have standing. The term originates from U.S. law and turns up in tech news almost every time a lawsuit fails before the actual substance of the case is even addressed.

Why many AI lawsuits fail at this hurdle

Hundreds of proceedings are currently underway around artificial intelligence. Authors, newspapers, music labels, and programmers are suing companies alleged to have used their works to train models. Many of these lawsuits are not dismissed because the courts consider the training to be permissible. They are dismissed because the plaintiffs failed to demonstrate that they personally suffered a concrete injury.

For companies, this is a very effective line of defense. A dismissal for lack of standing costs less time and money than a years-long trial over copyright. For the public, this has an uncomfortable consequence: fundamental questions remain unresolved. Whether a particular training method is legal is still unknown after such a ruling.

Standing therefore also helps determine which rules for AI ever come into being. Courts create law only where they rule on the merits. If plaintiffs fail en masse at the admissibility stage, the clarification shifts into politics — or never happens at all.

How courts measure standing

In the U.S., courts typically examine three points. First: is there a concrete, not merely hypothetical, injury? Second: was this injury actually caused by the defendant? Third: can a ruling actually remedy the problem, for instance through money or an injunction? If any one of these three points is missing, the lawsuit is inadmissible.

The first point is especially difficult when it comes to data. If a language model was trained on a novel, the novel still exists. So the author must explain what exactly their harm consists of. Lost licensing fees are one argument, lost revenue another. Mere outrage is not enough, even if it’s understandable.

A common misconception is confusing standing with winning or losing. Standing says nothing about who is in the right. It is merely the ticket of admission to the actual proceedings. You can have standing and still lose, and you can be right on the merits and still be turned away at the door. In Germany, there is a similar hurdle called Klagebefugnis, though it is structured differently.

Reading standing correctly in headlines

When you read that a lawsuit against an AI provider has been dismissed, a second look is worthwhile. If it says something like dismissed for lack of standing or inadmissible, the court did not evaluate the dispute at all. If, on the other hand, it says the use was lawful, that is a genuine ruling on the merits. These are very different pieces of news, even though both sound like a win for the company.

Often, such a dismissal is also not final. Courts frequently give plaintiffs the opportunity to amend their complaint and describe the harm more precisely. Many proceedings against major AI providers have continued this way in a second or third amended version.

Beyond courtrooms, you also encounter this term in data protection and consumer protection. After a data breach, the question regularly arises whether those affected may already sue, or only once their data has actually been misused. In class actions, too, standing determines who belongs to the group. Anyone who understands this hurdle better understands why some outrageous cases have no consequences in court.