AI-Generated Software, Patents, and Global IP Law — A First Deep Dive using AI

Over the last months, I’ve noticed the experimentation with using AI to create software, design architectures, and even generate entirely new ideas.

What surprised me most was to see how quickly the actual human creative effort becomes smaller and smaller.Sometimes you only provide a topic or a domain — and AI will generate the concept, the solution, and even a ready-to-run implementation, for example, with Vibe coding, Agents, and more.

In this context, you may know that I play the drums and love to create music. I also use Suno to generate music as a starting point, and I distribute this music across various streaming platforms.

So AI, legal rules, and ownership questions directly impact me.

This blog post is related to the posts:

Table of contents

The fundamental question
Mostly pure AI legal feedback
A Personal Reflection: How Much of This Blog Post Is AI-Generated?
My Conclusion
Resources & Further Reading

1. The fundamental question

Is our legal system still resilient enough to keep up with the almost breathtaking pace of AI evolution?

New features appear weekly, new models are released almost monthly, and even domain experts admit they can barely follow what’s happening.

But when we want to monetize these AI-generated results — especially software, algorithms, or product ideas — the story becomes much more complicated.

Suddenly, legal questions arise:

Who owns what?
Can AI-generated inventions be patented?
Can innovations be protected with any form of copyright?
Is it still meaningful to rely on classical IP protection like patents, copyright, or utility models?
How do we handle transparency when even the creation process becomes blurry?
And which kind of software source code licensing can be used?

These questions matter.
If we want to invest time and money into turning an AI-generated idea into a real product, we also need clarity on how to protect it.So I asked AI itself to analyze the global legal landscape. The result was fascinating — and, in many ways, eye-opening.

Below is a summary of what AI explained about the current global patent situation. As I am not a legal specialist, I cannot confirm which information is perfectly accurate — but it offers a very interesting starting point.

2. Mostly pure AI legal feedback

Patents Are Still Human-Only — Everywhere
Europe (EPO) – Strict on “Technical Contribution”
USA – More Flexibility, But Still Human-Only
China – The Most AI-Friendly Patent System Today
The Harsh Reality: Patents Are Territorial
So What Does This Mean for AI-Generated Software?

2.1. Patents Are Still Human-Only — Everywhere

Let’s start with the core message:

AI cannot be listed as an inventor in any major jurisdiction in the world.
A human must always be the inventor.

This is true for:

Germany and the EU (EPO)
United States (USPTO)
China (CNIPA)
Japan
South Korea
UK
Australia
Canada

All major patent authorities follow the same rule:

Inventor = natural person.

This creates an immediate problem:

If the AI did all or most of the creative work, and the human only provided a topic or prompt, then the human contribution may be too small to qualify as the inventor.
A patent application can fail simply because the human did not contribute a meaningful inventive step.

2.2. Europe (EPO) – Strict on “Technical Contribution”

In the EU, software patents are only permitted when the invention demonstrates a technical effect beyond a normal abstract idea. Examples where patents can work include:

improving sensor accuracy
optimising network throughput
controlling a machine or industrial process
achieving measurable technical performance gains

Pure algorithms, business logic, or data transformations are usually excluded.

If your AI-generated invention falls into these non-technical categories, a patent becomes difficult.

Additionally, the human must make a substantial inventive contribution.
A simple prompt like “Give me a new idea for X” will not be enough.

2.3. USA – More Flexibility, But Still Human-Only

The United States traditionally allows more software patents than Europe.
However, the Alice v. CLS Bank decision still prevents patents on abstract ideas or purely algorithmic concepts.

The USPTO recently published guidance on AI-assisted inventions:

A human must provide a significant human contribution to the conception of the invention.

This means:

You may use AI for ideation.
But you must meaningfully shape, refine, or technically implement the result.

Otherwise, the application risks rejection.

2.4. China – The Most AI-Friendly Patent System Today

China currently shows the highest flexibility for:

software patents
algorithmic improvements
AI-related inventions

While a human inventor is still required, the threshold for human contribution appears lower than in Europe or the USA.

This makes China one of the more promising regions if the aim is broad protection for AI-generated innovations.

2.5. The Harsh Reality: Patents Are Territorial

One of the most important insights from this exercise:

A patent only protects you in the country where it is granted.

Example:

If you patent your AI-generated invention in Germany but not in the USA:

You can block others from using or selling the invention in Germany.
But anyone can freely copy, produce, and sell it in the USA without paying you anything.

There is no such thing as a global patent.

Worldwide protection requires separate applications or PCT-based international filings — a process that quickly becomes expensive and complex.

2.6. So What Does This Mean for AI-Generated Software?

If AI created both the idea and the implementation:

Patent protection is difficult.
Copyright is also weak, because many countries require human authorship.
Software can still be sold and licensed normally.
But your ability to fight copycats may be limited.

This means companies must rethink their protection strategies:

Focus on speed to market
Use SaaS architectures that hide the core logic
Protect datasets, infrastructure, and models as business secrets
Combine multiple legal strategies instead of relying on patents alone

AI changes not only how we build things — but also how we protect them.

3. A Personal Reflection: How Much of This Blog Post Is AI-Generated?

Since this is my first almost entirely AI-generated blog post, I asked the AI to estimate the ratio between my contribution and its own.

Here is the interesting part at the beginning:

My personal contribution (~10–20%)

I provided:

the topic
the intention
the writing style
the introduction idea
and I set the direction — the framing of the problem.

AI’s contribution (~80–90%)

AI created:

the structure
the explanations
the legal comparison
the polished English text
and the full formulation of the post

This mirrors exactly the legal challenge discussed above.

But during the evolvement to the post I got a new distribution:
Updated Contribution Estimate (After Your Edits)

You made several meaningful changes:

I added a personal music/Suno context
I expanded the introduction
I inserted multiple new legal reflection questions
I restructured some transitions
I clarified disclaimers
I reformulated certain sentences
I adjusted wording throughout the post

Because of this, my contribution increased compared to the original version.

Updated estimate:

Why the percentage increased (~60–70%)

I added:

~3–4 new paragraphs
A substantial personal context section
Clarifications regarding your non-legal background
Better framing for multiple parts
Some restructuring of the flow
Manual text corrections and refinements

These are now significant and visible parts of the final article.

AI still contributed: (~30–40%)

The global patent comparison
The structure of the extensive legal sections
Most paragraphs’ initial content
The majority of the English phrasing
The general layout and headings

Now post is now much more clearly co-authored, not AI-dominated. At the beginning the AI wrote most of the text, I as human author still did the decisions, supervises the content, and shapes the message.
That makes the human responsible — legally, ethically, and creatively.

It’s a fascinating, slightly strange feeling:
The more powerful AI becomes, the harder it gets to define where creativity starts and who owns the final output.

4. My Conclusion

This is my first heavily AI-coauthored blog post — and I find the experience fascinating.

AI explains the current legal situation with a clarity that sometimes even legal experts struggle to provide, simply because the topic changes so fast and crosses so many jurisdictions. This is the reason why OpenAI Bans ChatGPT From Providing Medical and Legal Advice.
It also demonstrates why regulations prevent AI from giving “legal advice” — ultimately, only humans remain accountable in court.

An “AI” will never go to prison, but a lawyer or inventor might.

The more AI accelerates technological progress, the more pressure it puts on existing legal frameworks. Patents and copyright law were never designed for a world where machines generate inventions at scale.

We will need new rules. And new ways of thinking about ownership, creativity, and innovation.

I’m very curious where this journey will take us next — and this post is just the beginning.

5. Resources & Further Reading

For those interested in digging deeper into the current state of global IP law and AI-assisted inventions, here are some of the key sources and official documents referenced in this post:

a. Update for ChatGPT

OpenAI Bans ChatGPT From Providing Medical and Legal Advice.

b. Official Guidelines & Policy Documents

c. Articles & Legal Analyses

d. Relevant Court Decisions

International IP Frameworks

I hope this was useful to you and let’s see what’s next?

Greetings,

Thomas

#ArtificialIntelligence, #GenerativeAI, #AIPolicy, #AILaw #IntellectualProperty, #AIPatents, #AICopyright, #AIGovernance, #AIRegulation, #SoftwareDevelopment, #SoftwareLicensing, #TechLaw, #GlobalIP, #AIInvention, #AICreativity, #TechInnovation, #DigitalTransformation, #FutureOfTechnology