Closed Source

Every computer program is written by people in a kind of writing language called source code. This text is the complete blueprint: whoever has it understands how the program works and can modify it. With Closed Source, the manufacturer keeps this blueprint under lock and key. Customers only get the finished, executable version, from which the blueprint can practically not be recovered. You are allowed to use the program, but neither look inside it nor pass it on. The opposite is called Open Source: there, the source code is public, and anyone may read it and further develop it.

What the manufacturer protects with it

Developing software costs a lot of money, whereas copying it costs almost nothing. A company that publishes its source code thereby gives away years of work. A competitor could rebuild the program and sell it more cheaply. Secrecy is therefore, for many companies, the foundation of their business model.

In the AI industry, the dispute over this is especially visible. The best-known chatbots from major US corporations are Closed Source: you can use them via a website or a programming interface and pay per request. How the model is exactly built and with which texts it was trained remains unknown. Other providers release their models for download so that as many developers as possible can build on them. Both are deliberate strategies, not technical necessities.

For users, this closedness comes at a price. Anyone who builds a company on such a system is dependent on its operator. If the price rises, the model’s behavior changes, or the service is shut down, there is hardly any alternative left. Experts call this dependency the lock-in effect.

From source code to delivered file

Before being sold, the source code is converted into machine language, a process called compiling. What remains are columns of numbers that the processor understands but a human hardly can. Comments, names, and the developers' train of thought disappear in the process. This exact file is what the customer receives.

You can imagine it like a restaurant dish. You get the finished plate and are allowed to eat it. The recipe with the exact quantities stays in the kitchen. With a lot of effort you can taste out some of it, but you certainly can’t recreate it reliably.

With AI models, the situation is somewhat different. Their behavior is not stored only in program code, but above all in billions of learned numerical values, the parameters. A model is considered open if these values may be downloaded. Closed Source here means: the parameters reside exclusively on the provider’s servers, and you only ever communicate with them from the outside. Additionally, license agreements and copyright law secure what would technically be possible.

Closed Source in everyday life and in business news

Most of the programs you use daily are closed. Windows, macOS, iOS, Microsoft Office, WhatsApp, Instagram, and almost all video games belong to this category. This only becomes noticeable when something doesn’t fit: an annoying feature can’t be removed, an old bug can’t be fixed yourself. You can only wait until the manufacturer delivers an update.

In financial news, the term comes up when discussing the valuations of technology companies. Closed software allows for stable subscription revenue and high profit margins, because no one can simply copy the product. Investors value this. At the same time, regulators are increasingly asking whether outsiders can even check closed AI systems for fairness and safety.

A common misconception: Closed Source does not automatically mean expensive, and Open Source does not automatically mean free. Many closed apps are free and earn money through advertising. Conversely, companies sell paid maintenance around open software. Open or closed describes the rights to the code, not the price.