inferwire
/
Opinion·3 min read

Why AI-Generated Code Should Be Open Source

Developer Jakub T. Jankiewicz argues AI-generated code should be open source to prevent license laundering and address copyright limits on machine output.

TL;DR

  • Software developer Jakub T. Jankiewicz argues all AI-generated code should be published under Open Source or Creative Commons Zero licenses [^1].
  • The proposal targets license laundering from open training data and the legal lack of human authorship in automated code [^1].

Background

Open-source software relies on licensing frameworks that mandate attribution or grant rights to modify and redistribute code. Large language models train on vast repositories of public code, including open-source projects and technical community forums. As software developers integrate automated AI assistants into daily workflows, questions emerge regarding who legally owns machine-generated code and whether proprietary software can ethically incorporate output built on shared human knowledge.

What happened

Blogger Jakub T. Jankiewicz proposed that all code generated with help from AI should be released under Open Source licenses or dedicated to the public domain using Creative Commons Zero (CC0) [^1]. Jankiewicz argues that AI models were trained by scraping open-source code written since RMS announced the GNU Project and the Free Software Movement, doing so without permission or proper attribution [^1]. Additionally, training datasets included posts from Stack Overflow and Stack Exchange sister projects, which operate under Creative Commons licenses that legally require attribution [^1]. Converting this shared training corpus into closed-source, proprietary software constitutes what Jankiewicz calls "license laundering" [^1].

Beyond training data, Jankiewicz highlights legal principles surrounding software authorship [^1]. Under U.S. legal logic rooted in the Monkey Selfie dispute, non-humans cannot hold copyrights—a precedent later cited by the U.S. Copyright Office regarding AI [^1]. Similarly, Wikimedia Commons requires purely AI-generated images to be tagged as Public Domain [^1]. However, copyright laws differ internationally; for example, countries like the UK grant limited copyright protection to computer-generated works [^1].

To ensure fairness across jurisdictions and honor the open-source ecosystem, Jankiewicz urges developers and companies to explicitly adopt CC0 or open-source licenses for AI outputs [^1]. The argument applies particularly to workflows where developers rely on AI without reviewing or modifying the output [^1]. When human effort is limited primarily to writing prompts, Jankiewicz asserts the resulting code should unequivocally be open source [^1].

Why it matters

The proposal addresses a growing ethical and legal friction point in modern software engineering. As automated tools generate larger portions of commercial codebases, companies risk building proprietary products on software outputs derived from open-source work without adhering to the original license terms [^1]. Framing this process as license laundering highlights how closed-source AI code generation can erode the open software ecosystem that trained the underlying models [^1].

The legal distinction between human authorship and automated output also creates compliance risks across international borders [^1]. Because jurisdictions like the U.S. and the UK treat computer-generated copyright differently [^1], vendors relying on unedited AI code may face uncertain intellectual property rights. Explicitly placing prompt-generated code under public domain or open licenses creates a transparent legal baseline while ensuring that public knowledge repositories continue to benefit from automated software generation [^1].

Practical example

Say Liam works as a software developer at a mid-sized cloud analytics firm.

While building a new backend API, Liam relies heavily on an AI coding assistant to output data parsing functions. Liam enters prompt instructions and accepts the generated code directly into the company repository without reviewing or modifying the logic [^1].

Months later, Liam's team audits the intellectual property of their codebase. Under U.S. copyright principles derived from decisions like the Monkey Selfie dispute, non-humans cannot hold copyright [^1], leaving Liam's unedited AI output without clear proprietary copyright protection. Furthermore, because the AI model trained on open-source projects and Stack Exchange platforms without attribution [^1], deploying the code inside a closed commercial product raises concerns of license laundering [^1].

To resolve the legal ambiguity and address regional differences like the UK's computer-generated copyright rules [^1], Liam's engineering group adopts a policy aligned with Jankiewicz's recommendation [^1]. The firm explicitly applies a Creative Commons Zero (CC0) license to the prompt-generated utility functions [^1], returning the machine-written code to the open ecosystem that made the AI model possible.

Related gear

We recommend Eric S. Raymond's foundational book on open-source development to understand the collaborative ecosystem that trained modern code models.

AdvertisementAmazon

The Cathedral & the Bazaar

Sources

  1. [1]Jakub T. Jankiewicz Blog — All AI-generated code should be Open Source
  2. [2]Hacker News — All AI-generated code should be Open Source