If you write code, you already have a copyright on it. You did not register anything and you did not add a notice. It happened the moment you saved the file.
That part is simple. What confuses people is how little of your code copyright actually covers, and what it does not stop a competitor from doing.
To be clear, I am not a copyright expert, and I am not a full time developer either. I am somewhere in between, which is probably why people keep asking me. Colleagues and friends who write code bring up the same questions over and over, so I decided to write the answers down. Where it gets properly legal, I link to Conrad Fahrenkrug, who is the one here who actually knows copyright.
What you own the moment you write code
Copyright treats a computer program as a literary work. That sounds strange at first, but the logic is simple: code is written expression, saved in a form someone can read or copy.
Two things have to be true:
- Originality. You wrote it, and it is not the only possible way to write that function.
- Fixation. It exists somewhere: a file, a repo, a build.
Once both are true, protection is automatic in most of the world. No form, no fee, no waiting. If you want the basics first, start with What is Copyright?.
What gets covered:
- Source code, in any language.
- Object code, the compiled version, with the same protection.
- Comments and documentation you wrote alongside it.
- Structure, sequence and organisation, meaning how you arranged modules and designed the flow, in the parts where those choices were creative rather than forced.
That last one is where most of the litigation happens. Copying your architecture without copying a single line can still be infringement, but only in the parts where you had real choices.
What copyright does not protect
Here is where developers usually get surprised.
Copyright protects how you expressed something, never the thing itself. For software, that leaves the following outside:
- Algorithms. The method is an idea. Your implementation is protected, the method is not.
- Functionality. What the program does.
- Ideas, procedures and mathematical concepts. Excluded by the treaties themselves, not just by case law.
- Anything the problem dictates. If there is only one sensible way to write it, or the format or hardware forces it, it is out. Same for the standard idioms everyone uses.
What this means in practice: a competitor can look at your product, work out exactly what it does, and rebuild it from scratch with their own code. That is legal. Copyright gives you no monopoly on the function, only on your text.
US courts use a test from Computer Associates v. Altai (1992) called abstraction, filtration and comparison. They break the program into layers, filter out everything unprotectable, and compare only what is left. A lot gets filtered out.
Two decisions worth knowing:
- Lotus v. Borland (1995): a menu command hierarchy counted as a method of operation, so it was not protected.
- Google v. Oracle (2021): the US Supreme Court held that Google's reuse of Java API declaring code was fair use. It assumed the code was copyrightable instead of deciding the point, so API copyrightability is still technically open.
If what you need is to stop others from using the method itself, copyright is the wrong tool. There is a comparison further down.
The laws that actually apply to software
Short version: this is one of the most harmonised areas of IP there is.
International
- The Berne Convention removed formalities, so protection exists without registration in every member country, and there are more than 180 of them.
- TRIPS, article 10, says computer programs are protected as literary works, whether in source or object code.
- The WIPO Copyright Treaty, article 4, says the same and makes clear it applies whatever the mode or form of expression.
United States
Software was not obviously copyrightable at the start. After the CONTU report, Congress amended the Copyright Act in 1980 to define a computer program in 17 U.S.C. §101, and added §117, which lets the owner of a copy make a backup and make the copies needed simply to run the software. This is what people are looking for when they search for the computer software copyright act: there is no separate act, it is an amendment to the Copyright Act of 1976.
European Union
Directive 2009/24/EC covers computer programs specifically, and gives lawful users rights the licence cannot take away:
- one back-up copy where it is needed,
- the right to observe, study and test the program to work out the ideas behind it,
- decompilation, where it is indispensable to achieve interoperability with another program.
That third point matters more than most developers realise. Reverse engineering for interoperability is legal in the EU, within limits, even if the EULA says otherwise.
Who owns the code
Automatic protection tells you the code is protected. It does not tell you who owns it.
- Employees. In the US, code written by an employee within the scope of the job is a work made for hire, and the employer is the author from the start. In many civil law countries the developer is the author by default, and the company gets the economic rights through the employment contract or a written transfer. Check the local rule instead of assuming.
- Contractors and agencies. This is where things break. In the US, contractor work only counts as work made for hire if there is a written agreement and the work fits one of nine statutory categories. Software generally does not fit them. So with no written assignment, the contractor keeps the copyright and you have a licence at best.
- Open source dependencies. You own what you wrote. You do not own the libraries, and their licences travel with your build.
The rule is boring and it works: get the assignment in writing, before the work starts. A paid invoice does not transfer copyright.
If you want to see how ownership and registration work together, Conrad covers the process in How do you obtain a copyright?.
Code written with AI
This one comes up in almost every conversation I have now.
The US Copyright Office position is that copyright needs human authorship, and prompting is not authorship on its own. If you type a prompt and ship the output untouched, that output is not protected.
What is protected:
- the code you wrote yourself,
- your edits, fixes and refactors of the generated code,
- the selection and arrangement, meaning how you chose and combined the pieces into a working program.
In practice most real repositories are mixed. A file that started as a suggestion and then went through review, debugging and refactoring has human authorship in it. A file pasted in untouched does not.
Two habits worth picking up now:
- keep the commit history clean, because it is the best evidence of what a human actually did,
- if you plan to register in the US, be ready to identify AI-generated material, since the Copyright Office expects applicants to disclose it and to disclaim more than trivial AI content.
Conrad looks at the wider picture in AI and Copyright.
Should you register software copyright?
Legally you do not have to. Practically it depends on where you might have to enforce.
Register if the US matters to you. Registration is not what creates the right, but:
- you cannot file an infringement suit over a US work until the registration has gone through,
- statutory damages and legal fees are only available if you registered before the infringement started, or within three months of publication.
Without that you are left proving actual damages, which is slow and expensive.
The deposit problem. Registration normally means depositing code, and code is usually secret. The US Copyright Office has options for exactly this. The standard route is the first 25 and last 25 pages of source code, and you can block out trade secret material or deposit with portions redacted. You do not hand over the whole repository.
Other countries. Some have registries that are genuinely useful as proof of date and authorship, including several in Latin America and China. Others have none. If you have a specific market in mind, check that one rather than generalising.
How long it lasts. In the US and the EU, the life of the author plus 70 years. For US works made for hire, 95 years from publication or 120 from creation, whichever ends first. Either way, longer than the software will be relevant.
Copyright, patents and trade secrets
Copyright is one tool. On its own it is rarely enough.
- Copyright protects your actual code and its structure. Free, automatic, very long. Does not stop reimplementation.
- Patents can protect a technical method, so they do stop reimplementation. But they are expensive, slow, examined, country by country, and what is patentable varies a lot. The EU excludes programs as such, and the US limits abstract ideas.
- Trade secrets protect what you never publish: internal algorithms, training data, infrastructure detail. Free and potentially unlimited in time, but gone the moment it leaks. They only work with NDAs and access control behind them.
- Trademarks protect the name and the logo, not the code. Separate registration, country by country.
- Licences are what you use day to day. Your EULA or your open source licence sets the terms on which your copyright gets used.
Most software companies end up using four of the five.
If someone copies your code
Infringement usually comes down to two things: access and substantial similarity in the protected parts. Then the filtering happens, and everything that was never protectable drops out of the comparison.
What helps in practice:
- version control history with real timestamps,
- distinctive elements with no functional reason to exist, like unusual comments, naming or harmless quirks, because they are hard to explain as coincidence,
- your written assignments, so ownership is not the first thing you have to argue about,
- registration, where you have it.
The realistic routes, roughly in order: a takedown to the host or the platform, a cease and desist, then litigation. Most cases end well before the last one.
Frequently asked questions
Is my source code automatically protected by copyright?
Yes. In Berne Convention countries protection starts when the code is created and fixed, with no registration. Registration adds enforcement benefits in some countries, above all the US.
Can someone build the same functionality legally?
Yes, as long as they write their own code. Copyright covers your expression, not the function, so independent reimplementation is allowed.
Do I need to register my software copyright?
Not in order to have the right. But in the US you need a registration before you can sue over a US work, and before you can claim statutory damages and legal fees.
Who owns code written by a contractor?
Usually the contractor, unless there is a written assignment. Paying for the work is not enough in most systems, so put the transfer in the contract.
Is AI-generated code protected by copyright?
Not the parts produced by prompting alone. Your own code, your edits, and the way you selected and arranged everything can be protected.
How long does software copyright last?
The life of the author plus 70 years in the US and the EU. For US works made for hire, 95 years from publication or 120 from creation, whichever comes first.
Where to go from here
Copyright and software fit together fine, but only inside clear limits. You get protection for free, in most of the world, for a very long time, on the code you actually wrote. You do not get protection on the idea, and you do not get ownership just because you paid for the work.
If you want the registration handled, or you need it in more than one country, take a look at our Copyright Registration Services. One iGERENT specialist coordinates the filings with local counsel, with a fixed quote and clear timelines.
Prefer to ask a couple of questions first? Contact iGERENT for a free, no-obligation quote.
Disclaimer: This article is for general information only, not legal advice. Copyright rules vary by country and by the facts of each case.
English
