Discovery (US Civil Litigation)

When two parties dispute before a civil court in the US, the process doesn’t begin with the trial. Before that lies a long phase in which both sides must show each other their material. This phase is called Discovery. Either side may demand that the other hand over documents, emails, chat messages, and databases if they relate to the dispute. Employees can also be questioned under oath, often for hours and with a transcript being made. Anyone who withholds or deletes material risks harsh penalties from the court.

Why internal emails suddenly show up in the newspaper

Discovery is the reason why reports on major US lawsuits repeatedly feature internal messages from corporations. These quotes rarely come from whistleblowers. In most cases, the companies had to hand them over themselves during the proceedings. Once they end up in court records, they are often publicly accessible.

For companies, this is a significant risk, regardless of the outcome of the case. A careless sentence from an internal chat can be quoted for years. In the antitrust cases against major platforms, this happened again and again. Lawyers therefore say that every email should be written as if it might later end up on the judge’s desk.

Then there are the costs. Large proceedings involve millions of documents that must be reviewed and assessed. Discovery can therefore consume several million dollars before the case is even argued. Some defendants settle out of court for this reason alone: a settlement is cheaper than disclosure.

From mountains of files to document review by software

The process follows fixed steps. First, one side sends requests specifying which documents it wants to see. The company must then preserve all potentially relevant data and is no longer allowed to delete anything. This obligation is called a Litigation Hold and takes effect even before the lawsuit has actually been filed.

Then filtering takes place. Not everything must be disclosed: communications between the company and its lawyers are protected, and trade secrets can be placed under restrictions. For this purpose, a team goes through every document and classifies it. With millions of files, this can no longer be done by hand.

This is where technology comes in, under the keyword eDiscovery. Software pre-sorts files, detects duplicates, and suggests which documents are likely relevant. A method called Technology Assisted Review learns from the lawyers' decisions and applies them to the rest of the collection. US courts have explicitly accepted such methods since 2012. Final responsibility, however, still lies with the humans who sign off.

Discovery in AI lawsuits and the difference to Germany

In the current lawsuits surrounding artificial intelligence, Discovery plays a central role. When publishers or authors accuse an AI provider of using their texts for training without permission, they need evidence. This evidence lies exclusively with the provider: in lists of training data, in internal emails, in logs. This is precisely why such lawsuits are fiercely contested, often for months, before the actual legal question is even addressed.

It is important to distinguish this from German civil procedure. There is nothing comparable there. Anyone who files a lawsuit must generally provide their own evidence. There is no general obligation to open one’s own archive to the opposing party. German lawyers therefore critically refer to the exploratory approach of the US side as “Beweisausforschung” (evidence fishing).

This also has practical consequences for European companies. Anyone with a US subsidiary or doing business in the US can become subject to a Discovery proceeding. In such cases, the disclosure obligation can sometimes collide with European data protection law, which limits the transfer of personal data. A common misconception is also that Discovery is already the trial. It is only the preparation. Most US civil cases end in a settlement and never go before a jury.