If AI Writes It, Who Owns It?

The conversation around AI and copyright is everywhere. Did large language models train on copyrighted books? On open-source GitHub repositories? On paywalled news articles?

Those are critical questions, and courts around the world will be answering them for years to come. Lately, though, I’ve found myself thinking about something else. Not about how AI models are trained, but about what happens after they create something entirely new.

Let's Start with a Blog Post

Imagine asking ChatGPT to write a blog post. It nails the tone, the structure, and the message. You fire up your CMS, copy-paste the text, publish it, and move on.

A week later, someone copies your post, word for word.

Can you stop them?

The answer depends on where you live.

In the United States, copyright generally requires human authorship. If a piece of content is generated entirely by AI, it may not qualify for protection at all. In practical terms, that means anyone might be free to copy it without infringing on your IP. A "no" is very likely, then.

In Europe, the picture is complex: courts require a work to be the author’s own intellectual creation, strongly suggesting that meaningful human effort is required, even if the exact boundary line remains blurry. I wouldn't be surprised if the same "no" would apply here as well.

Yet even if the law eventually resolves every copyright nuance, I don't think it changes the ultimate direction we're heading.

Europe is taking a distinct regulatory path through the AI Act, introducing transparency requirements for AI-generated or manipulated content. That made me wonder: will Generated by AI become the nutritional label of the internet? A helpful marker of origin, or the internet equivalent of “I couldn’t be bothered”?

When the Artifact Is Easy to Recreate

Instead of asking who owns AI-generated content, we could ask: what happens when creating an alternative becomes almost effortless?

Imagine I ask an LLM to write an opinionated article about the future of software testing.

Tomorrow, you ask it to write a provocative piece on the exact same topic.

Our articles won't be identical. They’ll use different words, examples, and structures. Yet they may arrive at many of the same conclusions.

Nobody copied anyone, yet neither article is particularly unique anymore.

That isn't a copyright problem. It's a reproducibility problem.

Software and the Declining Value of the Artifact

For decades, software companies treated source code as the crown jewels: the core IP, the defensible moat, the asset competitors couldn't touch.

That made sense when software was expensive to create and trivial to copy.

AI changes that equation.

Suppose I spend three days crafting a custom Laravel middleware. Tomorrow, another developer prompts an LLM: "Build a Laravel middleware that solves this exact problem."

The implementation will differ. The outcome may not.

Nobody copied my code. Copyright still works. What changed is that recreating the capability became cheap.

That doesn't mean someone can vibe-code the next SAP or AutoCAD. Complex software contains years of domain knowledge, edge cases, integrations, constraints, and hard-won decisions.

But it does mean something important: the easier a piece of software is to describe, the less defensible its implementation becomes.

AI doesn't eliminate the value of software. It puts pressure on the value of the artifact itself.

Experience (Still) Matters

None of this makes software engineering trivial.

Give two junior developers and a staff engineer the same problem and the same AI tools. All three may produce working code.

The difference is who spots the hidden N+1 query, the race condition under peak load, or the shortcut that becomes crushing technical debt six months later.

Generating an answer is getting cheaper.

Knowing whether it's a good answer isn't.

AI can produce options. Experience provides judgment.

Where Intellectual Property Actually Lives

I don't think AI makes intellectual property disappear. I think it changes the balance between different kinds of intellectual property.

Source code will still matter. A well-designed architecture, a carefully engineered system, and a reliable codebase are really valuable assets.

But increasingly, the harder things to reproduce are everything behind them:

• The product decisions that define what should be built.
• The domain knowledge that explains why it matters.
• The UX choices that make it intuitive.
• The countless small choices that make software reliable, scalable, and pleasant to use.
• The architecture that ties everything together.
• The accumulated understanding of customers, edge cases, constraints, and trade-offs.

AI can generate a solution. It does not automatically know whether it is the right solution.

A developer with deep experience can look at the same output and see things that are invisible to others: a performance bottleneck waiting to happen, a security risk, an abstraction that will hurt the team six months from now.

The strongest software companies won't necessarily be the ones that write the most code.

They'll be the ones that understand the problem the deepest and are able to come up with the best solutions.