
Trade Secret
A trade secret is economically valuable information that a company deliberately does not disclose and actively protects. In the tech industry, this includes things like source code, formulas, customer lists, or details about how an AI system is built.
A trade secret is information that is economically valuable to a company and that the company intentionally keeps to itself. Typical examples include a formula, a design plan, a pricing calculation, or a list of key customers. For such information to legally qualify as a secret, it is not enough to simply not mention it. The company must visibly protect it, for example through passwords, locked rooms, or contracts that prohibit employees from disclosing it. Anyone who steals or leaks someone else’s trade secret can be held liable. In Germany, this has been regulated since 2019 by the Act on the Protection of Trade Secrets, known as the GeschGehG.
The value of keeping the competition in the dark
Many companies thrive on being able to do something others cannot. As soon as competitors learn the process, they can replicate it and the competitive edge is gone. That’s why secrecy is more valuable to some firms than any machine on the factory floor. The best-known example is the Coca-Cola formula, which has remained unpublished for over a hundred years.
A trade secret is the alternative to a patent. A patent grants the inventor a monopoly for around twenty years, but in exchange requires that the invention be publicly described. A trade secret never expires, but it only protects for as long as it actually remains secret. If someone else independently comes up with the same idea, they are allowed to use it. Companies therefore make a conscious choice: publish and get time-limited protection, or stay silent and hope for lasting protection.
In the AI industry, this question is particularly sensitive. Many providers treat as secret which data a language model has seen, how many computing components it has, and which tricks were used to train it. This is precisely what leads to conflicts, as regulators and researchers demand more transparency.
What a company must do for the protection to apply
Legal protection does not arise automatically. In the event of a dispute, a court examines whether the company has taken so-called reasonable secrecy measures. These include technical measures such as access rights, encryption, and separated networks. They also include organizational rules, such as ensuring that only a few people may view the complete documents.
The second component is contracts. Very common is the confidentiality agreement, known in English as a Non-Disclosure Agreement or NDA. In it, employees, suppliers, or investors commit not to pass on certain information. Violations can result in claims for damages. Importantly, there’s a flip side: those who protect their secrets carelessly lose in a dispute, even if the information was objectively valuable.
Not every form of imitation is prohibited. Anyone who takes apart a purchased product and learns from it is engaging in reverse engineering, which is generally permitted in Europe. Employees are also allowed to take their general experience and knowledge with them when they change jobs. What is prohibited is the theft of specific protected documents and data. In practice, the line between the two is often at the heart of court cases.
Trade secret disputes in tech headlines
In business news, trade secrets almost always appear in the context of a lawsuit. A well-known case was the dispute between Waymo, Google's self-driving car subsidiary, and the ride-hailing service Uber. An engineer had copied thousands of files on self-driving car laser technology when he switched employers. The case ended in 2018 with a settlement worth several hundred million dollars.
You encounter this principle in everyday life too. When an AI provider won’t reveal what texts it was trained on, it usually invokes trade secrets. When students or employees copy internal company documents into a public chatbot, this can inadvertently disclose a secret. This is precisely why many companies ban certain AI services in the workplace or run their own in-house versions.
A common misconception is equating trade secrets with data protection. Data protection protects individuals and their personal information. A trade secret protects a company’s economic interests. The two can overlap, for example in the case of a customer list, but they follow different rules and serve different purposes.