The growing legal argument that all AI-generated code must be open source
A new argument claims turning AI output into proprietary software is license laundering and legally unenforceable under US copyright law.
Policy · Source: Hacker News
What happened
Jakub T. Jankiewicz sparked a massive debate on Hacker News with a blunt and controversial claim. He argues that all AI-generated code must be strictly open source. The foundation of his argument rests on how modern AI models are actually trained. Companies scrape decades of open source code written since the early days of the GNU Project. They also ingest millions of answers from Stack Overflow and Stack Exchange. This data is used entirely without permission or proper attribution.
Jankiewicz calls the current system blatant license laundering. Developers take collective knowledge released under strict Creative Commons licenses and lock it behind proprietary products. He points out a massive legal flaw in closed-source AI code. Purely AI-generated code lacks human authorship. Under current US law, non-humans simply cannot hold copyrights.
The US Copyright Office has already established this precedent. They cite the famous monkey selfie dispute to deny copyright protection to AI outputs. Wikimedia Commons follows this exact logic by forcing all purely AI-generated images into the public domain. Jankiewicz argues that modern software development is heading the same way. Developers increasingly rely on AI tools without even reviewing the generated output. Code generated mostly through prompts with zero human modification belongs in the public domain.
Key facts
- Creative Commons — License used by Stack Overflow that legally requires attribution
- Monkey Selfie dispute — Legal precedent establishing that non-humans cannot hold copyrights
- UK — Country that grants limited copyright protection to computer-generated works
- CC0 — Creative Commons Zero license recommended for dedicating AI code to the public domain
Why it matters
Founders building proprietary software with heavy AI assistance face a massive legal gray area. If you rely on AI to write your core logic without significant manual modification, your codebase might not be legally yours. Competitors could theoretically copy your AI-generated modules without violating US copyright law. You cannot defend what you do not legally own. Investors will eventually catch on to this risk. Due diligence processes will start requiring deep audits of how much AI was used to build your proprietary tech stack. If your moat is just a wrapper around public domain code, your valuation will tank.
This dynamic pushes the software industry toward a dangerous split based entirely on geography. The UK currently grants limited copyright protection to computer-generated works, unlike the US. We might see startups moving their intellectual property holding companies to jurisdictions with friendlier AI copyright laws. Meanwhile, open source purists will increasingly weaponize these legal ambiguities. They will pressure companies to open their codebases or face public backlash. The definition of what constitutes enough human effort to earn copyright protection will become the most important legal battle in tech.
For builders
Audit your AI code generation tools today
Track exactly which parts of your codebase are purely AI-generated versus heavily human-modified. If you face an intellectual property audit during fundraising, you must prove human authorship to secure your valuation. Founders who blindly copy-paste AI output risk failing legal due diligence and losing investor trust.
Understand international copyright law differences
US copyright law currently denies protection to non-human creators based on established legal precedent. However, countries like the UK offer limited protections for computer-generated works. Global founders must decide where to incorporate their intellectual property holding companies based on how much AI code they use.
Add undeniable human value to code
Do not rely on raw AI output for your core proprietary business logic. You must review, modify, and heavily adapt the generated code to fit your specific architecture. This manual effort is the only way to legally transform public domain AI output into copyrightable proprietary software.
My take
I build AI products in public, and I know exactly how much we all rely on generated code. Calling it license laundering is harsh but technically accurate. If you are just writing prompts and shipping the raw output without heavy modification, you do not own that code. Stop pretending you are a genius developer and start protecting your business by adding actual human value to your products. The legal system will not protect lazy founders who try to copyright a machine's work.
Original reporting: Hacker News. This is my rewrite and opinion.