Should AI-written code be shared? One proposal opens a larger argument
open source
Software shared so people can inspect, change, and redistribute its source code under stated terms.
license laundering
A label for hiding or losing the sharing conditions attached to source material.
CC0
A public-domain dedication that lets a creator waive rights as far as law allows.
What happened
Jakub T. Jankiewicz, an open-source developer and blogger, argues in a personal post that AI-generated code should be open source. This is a proposal, not a new law or company policy. The post drew strong attention on Hacker News, a site for technology discussions. Points and comments measure attention, not whether the argument is correct.
The two arguments
Jankiewicz starts with the data behind AI models. He says they were trained on a huge body of open-source software. He also points to Stack Overflow, a programming question site, and related Stack Exchange sites. Some material there uses Creative Commons licenses that require attribution. In his view, turning shared knowledge into closed products resembles license laundering. The original sharing terms become hard to see, while the output becomes private. This is the author’s fairness argument. The post does not prove which files any particular model used.
His second argument concerns copyright. He says code produced entirely by a machine may lack a human author. It may therefore fail to qualify for copyright protection. He refers to the monkey-selfie debate and the US Copyright Office’s approach to human authorship. Laws differ between countries. He proposes that developers and companies release AI-generated code under an open-source license. They could also dedicate it to the public domain with CC0. He especially supports this when a person mainly writes prompts and does little review or editing.
Why the boundary matters
Real projects often mix human and AI work. A person may choose the goal, architecture, tests, and revisions. An AI system may write parts of the implementation. The post offers no shared test for deciding when that becomes pure AI output. Releasing everything could help others study and improve the code. It could also reveal a company’s private design or hard-earned knowledge. That tradeoff is why a simple rule affects more than licensing.
What the discussion adds
The Hacker News thread contains several challenges. Some comments ask why AI should be treated differently from programmers who learn from public code. Others compare AI with a compiler or another tool. Another question is whether an open-source license has much force for work that has no copyright owner. These comments do not settle the issue. They show that the important line is disputed. Hacker News activity is community interest, not evidence that the post is legally or technically right.
What is confirmed
The post clearly makes a moral and legal proposal. It connects training data, attribution, human authorship, open-source licenses, and CC0. It does not announce a court ruling or a policy adopted by a major project. The available material also does not establish what a particular model copied, whether a specific output matches training code, or whether a license condition was triggered.
What remains unclear
There is no agreed test for how much prompting, architecture, testing, or editing changes the result from pure AI output to human-authored work. The post does not settle how to separate private product design from generated code. It also does not show whether a training license condition automatically follows the generated result. Those questions depend on the facts of each project.
What to watch next
The next useful signals are practical. Watch whether projects ask developers to disclose AI involvement. Watch whether tools record the origin of generated code. Watch how courts and lawmakers describe human contribution. Until those answers become clearer, teams may need to record prompts, reviews, and edits before choosing a license. The value of this story is not that it proves one final answer. It makes the hidden boundary between human work and machine output easier to discuss.
Should code made by AI be shared?
📰 Full story: Should AI-written code be shared? One proposal opens a larger argument
A developer says AI-made code should be shared. Hacker News, a technology discussion site, talked about it widely.
open source
Code that people may read, change, and share under stated rules.
CC0
A way to offer work for anyone to use.
Hacker News
A website where people discuss technology stories.
💡 The gist
- A developer proposes sharing code made mostly by AI.
- He says AI learned from code shared by many people.
- Hacker News attention does not prove the proposal is correct.
Jakub T. Jankiewicz is an open-source developer and blogger. He wrote the original post. Open source means people can read, change, and share source code under stated rules.
He gives two main reasons. First, he says AI models learned from large collections of public code. He mentions Stack Overflow, a programming question site, and related sites. Some posts use licenses that require credit. Jankiewicz says private products may hide those shared origins. He calls this license laundering.
Second, he questions copyright. Code made entirely by a machine may have no human author. It may then lack copyright protection. He points to the monkey-selfie debate and US copyright guidance. Laws differ across countries. He suggests an open-source license or CC0. CC0 offers work to the public domain.
His idea is clearest when people only write prompts. It is less clear when people design, test, and edit code. Most software can mix human and AI work. No simple test separates these parts. This matters because sharing can help others learn. It can also reveal private product designs.
The Hacker News thread raised several questions. Human programmers also learn from public code. AI may be a tool, like a compiler. If code has no copyright, a license may work differently. These comments show a disputed boundary. Popular discussion shows interest, not truth.
The post does not show which files trained a particular model. It does not prove that an output copied existing code. It also does not settle the law everywhere.
Next, watch rules about AI disclosure. Watch tools that record code origins. Watch legal guidance about human contribution. Recording AI use, review, and edits can help later.
💬 Should people publish code made with AI?
The discussion asks whether using AI should make a software project open source. Some commenters say it should because models learned from open code; others say using a tool does not remove the creator's choice to keep a project private.
- Some people say AI-only code may not have copyright, so a license may be weak. Others say an open-source license still sets rules for copying, changing, sharing, and responsibility.
- Publishing the source can still be useful because people can inspect the program. But commenters disagree about how much value unreviewed AI code has compared with code that humans test and maintain.
- One side says AI learned from open-source projects, so its output should follow similar sharing rules. The other side says humans also learn from public code without publishing everything they create.
- GPL and AGPL raise a special question: does AI output based on them have to use the same kind of license? The comments say the law does not clearly answer that yet.
- Human steering, editing, and testing may affect copyright, but effort alone may not be enough. US legal authorities and courts have not produced a final answer.
- Private use is different from distributing software. Many commenters say the main licensing issue begins when the program is given or sold to other people.
initial digest at 281 comments (revision 1). We fetched 100 comments and sampled 100 across the thread. These are HN users’ reports, not independently verified facts.
Should AI-made code be shared?
📰 Full story: Should AI-written code be shared? One proposal opens a larger argument
One developer wants AI-made code shared with everyone.
open source
Code that everyone can see and change.
CC0
A sign that lets anyone use the work.
Hacker News
A website where people talk about technology.
AI writes instructions for computers.
Jakub T. Jankiewicz is a developer who wrote about this idea.
He says AI-made code should be open source.
Open source means people can see and change the code.
He says AI learned from code people shared.
So private code can feel unfair to him.
CC0 means anyone may use the work.
But people also plan, test, and fix code.
That makes the boundary hard to see.
Hacker News, a technology talk website, discussed the idea widely.
Popularity shows interest, not truth.
People should record which parts AI helped create.
💬 Should everyone see a computer's code?
People are asking whether a program made with AI should be shown to everyone.
- Some say yes because AI learned from public programs. Others say people learn from books and code too, but do not have to publish everything they make.
- Showing the code lets people look inside and check it. But the law may not give every AI-made program the same protection, and the rules are still unclear.
- People also disagree about whether careful human work changes the answer. A secret program used only by its maker is different from a program given to other people.
initial digest at 281 comments (revision 1). We fetched 100 comments and sampled 100 across the thread. These are HN users’ reports, not independently verified facts.
💬 Should all AI-generated code be open source?
HN commenters split over whether AI-generated code should be released publicly because AI models learned from open-source software. The debate covers copyright, license enforcement, human contribution, private use, and the still-uncertain state of US law.
initial digest at 281 comments (revision 1). We fetched 100 comments and sampled 100 across the thread. These are HN users’ reports, not independently verified facts.