The intersection of open source development and artificial intelligence (AI) is currently a minefield of valid excitement and legitimate anxiety. Recently, the Fedora Project introduced the Fedora AI-Assisted Contributions Policy. The full policy text outlines the basic ground rules. I was informally involved in the drafting process. Since its rollout, I am actively adjusting my own workflow to comply with its requirements.
For those outside the Fedora ecosystem, the policy establishes basic ground rules for how AI can be used in the project. It operates on three main pillars:
Accountability: You can use AI, but you own the output. The human contributor is always the author and fully accountable for the submission’s quality, license compliance, and utility.
Transparency: If a significant part of a contribution is taken unchanged from an AI tool, you must disclose it (typically via an
Assisted-by:commit trailer).Evaluation Limits: AI cannot act as the final judge on substantive contributions or evaluate a person’s standing within the community.
As someone who relies on these tools, I want to share my perspective on why this policy matters, the harsh realities of enforcing it, and why we must tactically embrace AI to protect the future of software freedom.
The "Accountability" Reality and the Threat of "Slop" 🔗
On paper, the Accountability clause looks like a strong deterrent against low-quality, automated spam. In reality, it functions primarily as a legal liability shield. If a contributor breaks a system by submitting AI-generated falsehoods, the policy simply puts the fault at their feet.
It does not, however, stop the spam.
We saw this clearly during a recent round of the Outreachy internship program. We had a record number of contributions recorded, but very few were actually merged. The volume of noise and sloppy contributions ballooned compared to past rounds. Many newcomers ignored the transparency mandate and flooded our git repositories with low-quality contributions. What actually deterred this "AI slop" wasn’t the text of the policy; it was the tangible enforcement mechanism of internship eligibility.
The primary consequence of submitting AI spam isn’t legal prosecution. It is reputational damage.
The Ethical Elephant in the Room 🔗
This brings us to the loudest objection from the open source community: the ethics of AI generation. Many maintainers deeply feel that LLMs are fundamentally engaged in theft. Maintainers recognize that the companies behind the LLMs extracted immense value from 40+ years of open source material without contributing value back, bypassing licenses and author attribution entirely.
I sympathize with the maintainers who hold this objection. It is frustrating to watch value extracted without reciprocity. However, I do not share the same sense of existential panic as other maintainers.
My view goes that the old rules simply no longer apply. Open source projects must evolve to survive. Instead of playing defense and fighting a philosophical war we cannot win, we need to think about how to tactically use AI to advance the interests of software freedom in a time when it has never been more threatened.
The Fedora AI-Assisted Contributions Policy Dual Mandate 🔗
When used responsibly, AI has the potential to be a massive equalizer. It can lower the barrier to entry for non-native English speakers drafting documentation and help junior developers navigate legacy codebases with decades or more of context. Democratizing access is how we attract the next generation of open source contributors and strengthen the social fabric of our community. This includes Fedora!
But there is a dark side to this democratization. Dumping a massive volume of low-effort, AI-facilitated contributions onto the laps of Fedora maintainers is a recipe for disaster. Many Fedora contributors are already stretched thin, doing more than is required of them. Forcing them to review a tidal wave of "AI slop" breeds deep resentment; to them, it feels like the active "enshittification" of our own community.
This is why we cannot focus solely on the people side of AI inclusion. We must carefully balance it by providing maintainers with the resources, support, and AI-facilitated tooling they need to survive. Practical improvements need deeper research and evaluation, such as using AI for advanced automated testing, scaling review workflows, and complex code refactoring. We should use AI to improve the quality and efficiency of our community’s work, ensuring that the same technology that lowers the barrier to entry also helps maintainers manage the gates.
If we handle this correctly, AI won’t replace the community. It will empower a more inclusive, diverse, and heavily-fortified one. The Fedora AI-Assisted Contributions Policy is our first step. Review the official Fedora Project policy to see the framework in action. Let’s keep the process open, keep the declarations honest, and tactically build the future of open source together.