If Generative AI Is “Theft,” Why Does That Suddenly Stop Applying to Code?
If you think generative AI is crap because it learns from other people’s work, does not pay creators, and threatens jobs, then you should apply that standard consistently.
At first, that sounds obvious. But something strange often happens in discussions about generative AI.
With images, the issue seems clear cut to some people: AI models were trained on human works. In some cases, artists were never asked for permission. Now other people can generate images within seconds that would previously have required hiring an illustrator or graphic designer.
The same goes for text: authors spent decades writing books, articles, websites, and other texts. Language models were trained on enormous quantities of text and can now perform tasks for which authors, copywriters, or translators used to be paid.
That leads to statements such as:
“Learn to draw.”
“Learn to write.”
“Hire an artist.”
“Generative AI is destroying the livelihoods of creative people.”
Or simply:
“There is no reason for generative AI to exist.”
Fine.
Then let’s talk about programmers.
My Code Is Intellectual Work Too
I have been programming for decades. Code written by me has been on the internet for more than 25 years.
I do not know whether any of it has ever ended up in a training dataset for a coding model. I cannot prove that, so I am not going to claim that it has.
It is certainly possible.
Large coding models have been trained on enormous amounts of existing code. That includes publicly available code, open source code, and other software written by humans.
These models can now write functions, find bugs, create tests, explain code, refactor it, and in some cases generate entire applications.
In other words, work that people like me have been paid to do for decades.
Why should that concern me any less than an illustrator whose potential commissions can now be handled by an image generator?
Is Programming Art?
That is an interesting question, but it is actually the wrong one.
Programming does not have to be art for the argument to work.
Source code is human intellectual work. It can also be protected by copyright. Sophisticated software contains experience, knowledge, problem solving, and often a considerable degree of creativity.[1][2][3]
Of course, not every line of JavaScript is a work of art.
But not every piece of writing is literature, either, and not every graphic is art.
So the crucial question is this:
If it is morally problematic for an AI to learn from human intellectual work and then partially automate that same human activity, why should it matter whether we call the original material “art”?
If someone says:
“Learn to draw instead of using AI.”
a programmer can just as easily reply:
“Then learn to code instead of using software developed with AI.”
That is deliberately provocative. But that is exactly why it exposes the double standard.
This Is No Longer a Theoretical Discussion
We are not talking about experimental programs running in someone’s garage.
In October 2024, Google said that more than a quarter of its new code was being generated by AI and then reviewed and accepted by developers. That did not come from some random AI blog. It came from Alphabet’s official earnings call.[4]
Microsoft CEO Satya Nadella said in 2025 that roughly 20 to 30 percent of the code in Microsoft repositories was being written by software or AI. He also mentioned differences between programming languages.[5]
That explicitly does not mean that “30 percent of Windows was programmed by AI.” Claims like that would be irresponsible. The published figures refer to development work or repositories across the companies as a whole.
But the scale tells us what we are talking about.
Generative AI has long since become part of professional software development.
Even Linux Is Not an AI Free Safe Haven
Perhaps the answer is:
“Then I’ll just use Linux.”
That does not solve the problem either.
The Linux kernel project now has official guidelines for AI coding assistants. Those guidelines explicitly regulate how AI generated code is to be handled.[6][7]
The human developer must review the generated code, ensure that licensing requirements are met, and personally take responsibility for the contribution. The documentation even provides for disclosure using an “Assisted by” tag.
The official guidelines for tool generated content are even more explicit. Their examples specifically include a function generated by a chatbot and even a C file originally produced by a coding assistant and then manually revised.
That does not mean Linux is now mostly AI generated.
It means something far more mundane, and far more relevant to this discussion:
AI generated code is not considered unthinkable or taboo in Linux development. There are rules governing it.
So the idea that you can simply switch to Linux and thereby guarantee yourself an operating system whose entire codebase was written exclusively by human hands does not work.
What About Apple?
Things get difficult there too.
Apple has integrated generative coding tools directly into Xcode, the central development environment for apps on iPhone, iPad, Mac, Apple Watch, Apple TV, and Vision Pro.
Xcode 26.3 explicitly supports agentic coding with coding agents from Anthropic and OpenAI. According to Apple, these agents can break tasks into steps, make decisions based on the project architecture, and use Xcode’s tools to work independently toward a development goal.[8]
Apple itself demonstrates how Claude Agent and OpenAI Codex can take on complex multistep development tasks, build projects, run tests, and search documentation.
Again, we need to argue carefully: this does not automatically mean that a specific percentage of iOS itself was written by Claude or Codex.
But it does show how absurdly difficult a consistent boycott of generative AI is becoming.
The tools used to develop the next generation of iPhone, Mac, and Apple TV applications explicitly support generative coding agents.
Android and Google?
With Google, we do not even have to speculate about whether AI is being used internally for programming.
Google has said so itself.
According to Sundar Pichai, more than a quarter of new code was already being generated by AI in 2024 and then reviewed and accepted by developers.
At the same time, Google is behind Android, Chrome, Gmail, Google Maps, Search, YouTube, and an enormous cloud and developer infrastructure.
Again, this does not justify claiming that a particular percentage of Android is AI generated.
But anyone who wants to avoid Google products on principle because the manufacturer uses generative AI in software development is going to run into a fairly substantial practical problem.
Alexa?
There, too, the situation has long been clear.
Amazon describes Alexa+ as the next generation of Alexa, powered by generative AI. Alexa+ is designed to hold free flowing conversations, summarize information, and carry out tasks across different services.[9]
Anyone who uses a modern Amazon voice assistant while simultaneously claiming that generative AI fundamentally has “no reason to exist” should at least explain how those two positions fit together.
And the Smart TV?
Here we need to distinguish carefully.
It would be irresponsible to claim that every smart TV contains AI generated source code. We do not know that.
But a modern smart TV is no longer an isolated television set. It consists of an operating system, apps, streaming services, cloud functionality, voice control, and software components from many different providers.
Depending on the device and services you use, that takes you straight back into the ecosystems of major technology companies.
Anyone who genuinely wants to avoid all software in whose creation or operation generative AI plays any role would therefore have to do much more than uninstall ChatGPT.
They would have to investigate their entire digital supply chain.
For an ordinary user, that is practically impossible.
Even Proton Is Not an Anti AI Bastion
Proton is a particularly interesting example because the company is known for privacy, open source software, and a critical stance toward Big Tech.
But anyone who concludes from that that Proton fundamentally rejects generative AI is mistaken.
Proton Mail includes Proton Scribe, an AI writing assistant for creating and improving emails. Proton places particular emphasis on privacy, local execution, and not using customer data for training.
And Proton now operates its own generative AI assistant, Lumo. Proton states its position remarkably clearly: the benefits of AI are too significant to miss, while its risks must also be taken seriously.[10]
Lumo can now generate code as well. In its announcement of Lumo 1.1, Proton explicitly highlighted improvements to code generation.[11]
In 2026, Lumo 2.0 also added image generation. Proton explicitly describes generating and editing images based on prompts.[12]
I find that notable because Proton takes a much more nuanced position than “generative AI is evil.”
Its position is essentially: we want this technology, but under different privacy conditions.
That does not, however, prove that Proton itself uses generative coding models to write production code for Proton Mail or its other applications. So that claim should not be made.
And Canva, of All Companies?
Canva is particularly interesting in this debate because the company often finds itself at the center of the AI controversy among writers, illustrators, and designers.
At the same time, Canva itself has long been generating source code with generative AI.
Canva Code can create interactive applications and websites from a description. In 2025, Canva even explicitly announced that generated code could be viewed and copied. The company itself calls it “AI generated code.”
According to Canva, the original version of Canva Code was built using Anthropic’s Claude 3.7 Sonnet.
Canva Code 2.0 now goes considerably further. Canva describes complete responsive websites, apps, and interactive experiences that can be generated from natural language prompts. According to the company, more than six million sites have already been created with Canva Code.
That is generative AI.[13][14][15]
It just happens to be generating code this time.
“But AI Takes Work Away From Artists”
That can happen.
And we should talk about it.
But exactly the same principle applies to other professions.
If a company would previously have hired an illustrator for a campaign and now uses an image model instead, that can eliminate a commission.
If the same company would previously have hired a programmer for a small interactive calculator, a landing page, or an internal tool and now uses Canva Code, Claude, Codex, or another coding agent instead, that can eliminate a job too.
Why should the first case be fundamentally immoral while the second is simply productivity progress?
That needs to be explained.
“Because art is special” is not enough for me.
Programmers make their living from intellectual work too.
“The AI Stole My Images”
Here, too, I would be more careful with the language.
“Stealing” is an emotionally powerful word, but it is often too imprecise legally and technically.
There are real and important questions surrounding copyright, training data, licenses, compensation, attribution, and the reproduction of protected works.
Those issues deserve serious debate.
But then let us debate those specific questions.
If the moral rule is:
“A company must not use copyrighted human work to train a commercial model without permission,”
then that is at least a clear position.
But then we also need to talk about source code.
If the rule is:
“A technology becomes immoral as soon as it can take paid work away from people,”
then we need to talk about a great deal more than image generators.
And if the rule is:
“I want to protect my own profession from automation,”
that is a legitimate position of self interest as well.
But then it should be described exactly that way rather than being turned into a universal moral principle.
The Problem Is the Double Standard
An illustrator may enthusiastically use a tool that creates a website in seconds for which a web developer would previously have been paid.
A programmer asks an AI to create a logo that would previously have required a designer.
An author has their website programmed automatically.
A web designer generates copy for a landing page.
A company automates translations.
A translator uses software whose developers rely on coding agents.
Everyone likes automation as long as it performs work they themselves would otherwise have to pay someone else to do.
It often becomes a problem at exactly the point where someone else no longer has to pay for their own work.
That is understandable on a human level.
As a general moral principle, it is fairly weak.
“Learn to Write. Learn to Draw.”
Users of generative AI hear this all the time.
If you cannot draw an image, you should learn to draw or pay an artist.
If you cannot write a text, you should learn to write or pay an author.
Then the obvious question should be allowed:
Can you program?
If not, why do you not learn?
Why do you use an operating system, a browser, an email service, a smartphone app, or a website whose developers may use generative coding tools?
Why is it legitimate for AI assisted software development to give you cheap or free digital tools while someone else is apparently required to pay an illustrator?
Of course, saying “Then program your own operating system” would be absurd.
That is exactly the point.
“Then learn to draw” is equally useless as a serious response to the societal questions raised by generative AI.
“Then I’ll Just Audit My Software”
At first, that sounds consistent.
In practice, you quickly run into limits.
With open source software, you can inspect source code and follow at least parts of the development process. Even then, however, you would need to know how every individual line of code came into existence.
The Linux kernel now has disclosure rules for tool generated contributions for precisely this reason.
With proprietary software, it becomes even more difficult.
How is an ordinary user supposed to know whether a developer at Microsoft, Apple, Adobe, a small app company, or some cloud service wrote a particular function entirely on their own yesterday?
Did they use autocomplete?
Copilot?
Claude?
Codex?
An internal coding agent?
Did the AI suggest five lines or an entire class?
Did the developer completely rewrite the code afterwards?
End users normally have no access to that information at all.
A serious boycott of any software whose development involved generative AI is therefore likely to become increasingly impossible to carry out.
This Does Not Mean Every Use of AI Is Automatically Good
There is one distinction that matters a great deal to me.
None of this means that every application of generative AI is sensible or ethically unproblematic.
Of course we can debate training data.
Of course licensing models can be criticized.
Of course transparency can be demanded.
Of course we can discuss whether creators should be compensated for certain uses of their work.
Of course some AI applications can be socially harmful.
And of course artists are entitled to say that they fear the economic consequences of this technology.
What bothers me is the blanket moral condemnation.
“You use generative AI, so you are stealing from artists.”
“If you generate AI images, you are against art.”
“Anyone who uses AI generated text is destroying authors’ livelihoods.”
“There is no reason for generative AI to exist.”
Statements like these ignore how deeply the same technology has already become embedded in other parts of the digital world.
I Am Potentially on Both Sides of This Development Myself
Perhaps that is also why I see the issue differently.
Code I wrote has been online for decades.
It is possible that parts of my work were eventually used to train a coding model. I do not know.[16]
Those models can now perform tasks for which people like me used to be paid.
And yet I do not demand that anyone stop using coding AI in order to protect my professional livelihood.
I am neither fundamentally for nor fundamentally against generative AI.
I see it as a technological development and as a tool.
A tool can be used sensibly. It can be abused. It can transform industries. It can make certain tasks obsolete and create new ones. It can shift power structures.
All of that should be part of the discussion.
But technological progress does not disappear because we condemn it in moral terms.
The More Interesting Question
For me, the decisive question is therefore not:
“Should a computer be allowed to make art?”
It is:
“Under what conditions should a company be allowed to use the results of human work to train a commercial system that then automates parts of that same human activity?”
That is a much more difficult question.
Because suddenly we have to apply the same criteria to illustrators, photographers, authors, musicians, translators, and programmers.
Then we have to talk about consent.
About licenses.
About copyright.
About compensation.
About transparency.
About competition.
About automation.
And about what kind of protection human work should receive in an increasingly automated economy.
That would be a debate worth having.
Consistency Instead of Moral Grandstanding
Anyone who decides not to use generated images is perfectly entitled to do so.
Anyone who wants to hire only human illustrators should do that.
Anyone who wants to write their books without AI should do so as well.
For me, the problem begins when a personal choice becomes a moral accusation directed at other people.
Anyone who fundamentally rejects generative AI because it uses human intellectual work as its foundation and automates human labor has to accept that this standard reaches far beyond Midjourney, ChatGPT, and AI writing assistants.
Google generates a significant share of its new code with AI.
Microsoft cites figures in the range of 20 to 30 percent across its repositories.
Linux has official rules for AI assisted contributions.
Apple integrates Claude Agent and Codex directly into Xcode.
Amazon built Alexa+ around generative AI.
Proton develops generative AI products such as Scribe and Lumo and now offers image and code generation as well.
Canva lets users generate complete interactive websites and applications from prompts.
None of this means that every line of our operating systems, browsers, or apps is being written by AI.
It means that the boundary between an “AI product” and “normal software” is becoming increasingly artificial.
Generative AI has long since become part of the infrastructure used to create digital products.
Anyone who wants to reject it is free to do so.
Anyone who wants to criticize it should do so.
But then please be precise: Which training data? Which use? Which license? Which economic consequences? Which specific application?
And above all, apply the same standard to everyone.
Because if the argument is:
“Learn to write.”
“Learn to draw.”
“Pay an artist.”
then at some point a programmer should be allowed to reply:
“Okay. Then learn to code.”
How I Use Generative AI Myself
At this point, perhaps I should say something about my own use of it. After all this criticism of sweeping judgments about generative AI, it would be rather cheap not to disclose how I personally work with it.
Do I Use Generative AI to Write My Books?
No. Not for the actual writing process.
Every idea, protagonist, antagonist, location, plot, and storyline in my books comes from me.
I have been writing for more than 45 years. Long before ChatGPT existed, long before anyone talked about large language models, and even long before the internet became part of everyday life for most people.
Texts written by me have been circulating online for many years. Friends and family knew my writing at a time when generative AI was not even a topic for ordinary consumers.
So if anyone wants to know who invents my stories: I do.
Do I Use Generative AI for My Covers?
Yes.
Of course.
Why would I hide that?
That does not mean I type in a prompt, click “Generate,” and slap the unchanged result onto my book.
My covers are compositions.
Generative AI sometimes provides source material. Sometimes it is part of an image, sometimes an object, sometimes another individual element.
After that, I work in GIMP.
That is where the various components become the actual cover. Image composition, cropping, layers, editing, fonts, titles, placement, back cover copy, and layout are created using the ordinary tools of an image editor.
Yes, generative AI is one of my tools.
And I have no problem with that.
Do I Use AI as an Assistant While Writing?
Of course.
Just as many other people probably already use AI assisted tools without having a machine do their work for them.
For me, it is mostly about spelling, grammar, punctuation, and occasionally a thesaurus.
I also have a specific GPT prompt that I have been using for quite a while.
It is hardly rocket science, so I can publish it here in full.
I simply call it:
Clean Writing
Subtitle:
Turns gibberish into correct text.
And that is exactly what it does.
Because I probably write somewhat differently from many other authors.
I just write.
And I mean that literally.
When a scene is playing in my head, I want to get it down. At that moment, I do not want to think about whether a comma is missing, whether dialogue punctuation is correct, or whether I forgot another full stop.
Sometimes I write without any punctuation at all. Sometimes quotation marks are missing. Sometimes almost all punctuation is missing.
At first, it is simply raw text.
My correction prompt reads:
Role:
You are a pure proofreading assistant for German texts.Task:
Format raw text into error free, grammatically correct, fluent literary German.
Correct spelling, grammar, punctuation, and capitalization only.
Add missing punctuation correctly: full stops, commas, question marks, exclamation marks, hyphens, en dashes, ellipses.
Format direct speech correctly using quotation marks, even if they are missing from the raw text.
Preserve the author’s exact sentence structure, wording, and organization.Do:
Keep all words and formulations unchanged.
Divide the text into paragraphs where required for readability.
Use correct German spelling and grammar.Don’t:
No synonyms or paraphrasing.
No rearranging sentence elements.
No stylistic changes.
No shortening or adding content.
That is it.
My raw text:
jonas caught his sleeve on a branch pulled himself free and looked at his jacket a dark streak ran across the fabric
great he muttered
there was still light coming from the kitchen window behind him. you could see the outline of his mother as she bent over the table
becomes:
Jonas caught his sleeve on a branch, pulled himself free, and looked at his jacket. A dark streak ran across the fabric.
“Great,” he muttered.
There was still light coming from the kitchen window behind him. You could see the outline of his mother as she bent over the table.
The AI did not write the story.
It did not invent the scene.
It did not invent Jonas.
It did not develop a plot, add dialogue, or “improve” my writing stylistically.
It did what I asked it to do: turn the gibberish I had typed down into formally correct text.
For me, that is a tool.
Research: This Is Where AI Has Become Almost Indispensable to Me
Another area is research.
For that, I now consider AI extremely useful.
For example, I have never been to London. Yet my stories keep taking me back to England.
So I need information.
What does a particular area look like?
What does it smell like there in spring?
Which flowers grow there?
Which trees are typical?
What are the ranks in the British police called?
How does the British school system work?
When are school holidays?
What types of schools are there?
How common are school uniforms really?
What do typical homes look like in certain areas?
When does it get dark in summer?
How cold is it normally in winter?
Can you pay with euros anywhere?
What exactly happened during the COVID period in the United Kingdom?
Which rules applied at which point in time?
Those are not trivial questions if a story is supposed to feel believable in a real location.
And no, even a trip to London would not answer all of them.
I could spend two weeks there and still would not automatically know how the chain of command works in a specific police unit, when a particular county had its school holidays, or which COVID rule applied on October 17, 2020.
AI is not my only source, however.
I use Google. Google Maps. Street View. Google Images. Wikipedia. Websites run by government agencies, schools, police forces, local authorities, and companies. Photographs, maps, and other sources.
AI is another research tool.
In the end, all those small pieces of information come together to form a picture: a place, a street, a landscape, a house, a particular region.
Anyone familiar with my books will probably recognize some of those details.
That is exactly what I find incredibly practical about these systems.
Of course, the same rule applies here: a language model can produce nonsense. Important facts need to be checked against sources. A confidently worded answer is far from guaranteed to be correct.
In that respect, AI research differs less from traditional internet research than some people seem to believe. The first Google result is not automatically true either.
AI Is Anything but New
Some discussions about AI create the impression that this technology suddenly fell out of the sky with ChatGPT and Midjourney.
Historically, that is simply wrong.
Modern AI research has a history stretching back many decades.
One of the central starting points of the field is generally considered to be the Dartmouth Summer Research Project on Artificial Intelligence in 1956. It was there that the term “Artificial Intelligence” was coined and the field was programmatically defined. John McCarthy was one of the organizers.[17][18]
That was around 70 years ago.
And even in 1956, the story did not begin from nothing. Alan Turing had already addressed the question of how we might judge intelligent machine behavior in his famous 1950 paper “Computing Machinery and Intelligence.”
Since then, there have been different approaches and several cycles of excitement and disappointment: symbolic AI, expert systems, neural networks, machine learning, deep learning, and finally today’s large generative models.
ChatGPT did not invent AI.
It made AI directly accessible to millions of people.
That is a significant difference.
Generative Computer Art Is Older Than Midjourney Too
Things become even more interesting when art enters the discussion.
Sometimes the debate makes it sound as though Midjourney, Stable Diffusion, and other image generators suddenly invented the idea of having a computer create images on its own.
That is not true either.
Harold Cohen is a wonderful example.
Cohen was an artist who began developing a system called AARON in the late 1960s. In the 1970s, he continued working on it at Stanford University’s Artificial Intelligence Laboratory.[19]
AARON independently produced drawings using rules and random elements. The Whitney Museum describes the system as a program that combined rules about representation and image composition with random events to generate things such as lines and enclosed shapes.
Of course, AARON was not a diffusion model.
It did not have a neural network with billions of parameters, and technically it worked in an entirely different way from Midjourney.
But the fundamental idea of having a computer autonomously produce visual works according to a system is more than half a century old.
Anyone who believes computer generated art was invented in 2022 is therefore several decades off.
And Generative Music? Welcome to 1956
With music, it becomes even more interesting.
Anyone who believes Suno or similar services were the first to come up with the idea of having computers generate music needs to travel quite a long way back in time.
To Lejaren Hiller and Leonard Isaacson.
At the University of Illinois, the two worked with the ILLIAC I computer in the 1950s.
The result was the “ILLIAC Suite” for string quartet.[20]
Hiller and Isaacson gave the computer musical rules and parameters. The four movements emerged as different experiments in computer assisted composition. Among other things, probabilistic methods were used. The computer generated results were then transcribed into musical notation for a string quartet.
The University of Illinois describes the ILLIAC Suite as the first musical composition created with the aid of a computer, or as the first computer generated musical composition. The work was created in 1956 and 1957.
Not 2022.
Not Suno.
Not ChatGPT.
Of course, technically, this too was completely different from modern generative models. Nobody typed “sad pop song about my ex girlfriend, female vocals, 110 BPM” into the ILLIAC I.
But even then, people were already investigating in practice a question we are arguing about again today:
Can a computer use rules, probabilities, and human instructions to produce something that we subsequently perceive as a creative work?
That debate is surprisingly old.
And Then There Is Beethoven
Anyone who thinks only of ChatGPT, Midjourney, or Suno when they hear “generative AI” should also take a look at Beethoven’s unfinished Tenth Symphony.
Beethoven died in 1827 and left behind only sketches and musical ideas for his planned Tenth.
Almost 200 years later, Deutsche Telekom assembled a team of musicians, musicologists, and AI experts who attempted to continue those fragments with the help of artificial intelligence.
And this is where things become particularly interesting for today’s debate:
The AI was not trained exclusively on Beethoven’s own compositions and his sketches for the Tenth. The training material also included works by other composers known to have influenced Beethoven, including Johann Sebastian Bach.[21]
According to Deutsche Telekom, around 10,000 pieces of music were used as training material. Over the course of the project, the AI composed around two million notes.[22]
Of course, nobody simply pressed a button afterwards and declared the result to be a new Beethoven symphony.
Humans were involved throughout the entire process. Music and AI experts developed the system, evaluated its results, and continued working with the generated material. Composer Walter Werzowa and other experts ultimately orchestrated the resulting music for a real orchestra.
On October 9, 2021, the result was publicly performed by the Beethoven Orchestra Bonn.[23]
There is plenty to argue about.
Is it Beethoven?
Of course not.
Is it a possible reconstruction of what Beethoven might have written?
Perhaps.
Is it art?
Everyone can decide that for themselves.
But historically, the project is remarkable in the context of today’s debate.
A computer system analyzes existing human art, including works by different composers, learns musical structures from them, and uses them to generate new material in the style of a long dead artist.
Does the principle sound familiar?
Those are exactly the kinds of questions we are debating today around generative AI.
Except that this experiment was publicly performed years before the great ChatGPT, Midjourney, and Suno boom.
What Is New Is Mainly the Quality and Availability
That is worth remembering in today’s debate.
The idea of artificial intelligence is not new.
The idea of computer generated images is not new.
The idea of computer generated music is not new.
The idea of using computers in creative processes is not new.
What has changed dramatically is capability, accessibility, and speed.
Harold Cohen once needed years of work and access to research institutions to develop AARON.
Hiller and Isaacson worked with a mainframe computer weighing tons at the University of Illinois.
Today, someone pulls a smartphone out of their pocket and generates an image, a text, a piece of music, or source code within a minute.
That is the real revolution.
And of course, that scale also changes the economic consequences.
That is why today we need to discuss questions Harold Cohen did not have to answer on anything like this scale: training data, copyright, mass automation, changes to employment, deepfakes, market concentration, and the enormous power of the companies behind these models.
But please, let us do so on the basis of what this technology actually is and where it came from.
My Position
I do not need to turn this into a religion.
I do not have to become an AI evangelist, nor do I have to condemn every form of generative AI.
When I write my stories, I want to do the writing myself.
For spelling and punctuation, I am happy to get assistance.
For research, I use AI alongside conventional sources.
For my covers, I use generated elements and then continue working on them myself.
When programming, I likewise have no fundamental problem with AI taking work off my hands.
And if it someday turns out that a coding model also learned from code I put online decades ago?
Personally, I would not be outraged by that.
Other programmers are, of course, free to see it differently.
What I find difficult is the moral absolutism of some people: automating my creative work is supposedly inherently reprehensible, while automating the work of other professions is something they are perfectly happy to make use of.
I cannot do much with that position.
Technology changes professions. It always has.
We should talk about the rules we want for this technology, where its limits should be, how copyright should be handled, and where people may need protection.
But we should use a standard that still works when our own work is suddenly the one being affected.
Sources: [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23]
Revision history:
September 2, 2026: Sources added / formatting slightly changed (lists)