Luanti is free software. Anyone on earth can legally download it, copy it, modify it, rename it, and ship it. That is not a loophole, it is the entire point of its license.
Luanti was also just pulled from Google Play over a copyright complaint.
Both of those things are true at the same time, and the gap between them is where a lot of us are now living. The project, a voxel game engine that spent most of its life under the name Minetest, got taken down after a copyright notice that its maintainers describe as baseless. Not a nuanced dispute over a contested asset. Baseless.
What actually happened, in plain terms
App stores run on complaint machinery. Someone files a copyright claim, an automated system evaluates it, and if the claim clears whatever threshold the system uses, the app comes down. The developer then gets to argue their way back in.
That ordering matters more than anything else in this story. The removal comes first. The review comes second, if it comes at all. A project with no legal team and no revenue is expected to appeal into a process it cannot see, on a timeline it does not control.
For a project like Luanti, that is especially strange, because the accusation contradicts the software’s own licensing. Open-source code exists to be redistributed. Alleging that it infringes is a little like accusing someone of trespassing on a public park.
Where the AI part comes in
This is the piece I think non-technical readers most need to understand, because it changes how you should read every takedown story from here on.
An AI agent is not a chatbot you talk to. It is software given a goal, a set of tools, and permission to act without a human signing off on each step. Scan these listings. Compare them to this catalog. File a notice when the similarity score crosses a line. Repeat, forever, at a scale no legal department could match.
Notice what is missing from that loop. Nobody reads the license file. Nobody asks whether the thing being “copied” was explicitly built to be copied. Nobody weighs the cost of being wrong, because from the agent’s side, being wrong costs nothing at all.
Meanwhile the agent on the receiving end, the enforcement system inside the store, is often also automated. So you get two pieces of software negotiating the fate of a real project, and the first human to look closely at the situation is the developer reading a removal email.
The asymmetry nobody voted for
Here is what the setup looks like when you lay the two sides next to each other:
- Filing a claim is nearly free, instant, and can be repeated thousands of times a day by a script.
- Answering a claim takes a human being, time, patience, and sometimes legal advice, once per incident.
- Being wrong as a claimant usually carries no consequence.
- Being wrong as a developer means your app is gone from the only storefront most Android users will ever open.
You do not need bad intent anywhere in that system for it to produce bad outcomes. You just need speed on one side and paperwork on the other.
Why this is not only an open-source problem
Luanti is a sympathetic case because the contradiction is so obvious. Free software, accused of theft. That clarity makes good headlines.
But the same machinery points at everyone. A small studio, a solo developer, a nonprofit shipping an accessibility tool, a teacher publishing a classroom app. None of them have a rapid-response team. All of them are one confident, automated accusation away from vanishing from a store they depend on.
Open-source projects are the canary here mostly because they publish everything in the open, so when the system misfires, we can all see it clearly.
What good automation would look like
I am not arguing that platforms should review every complaint by hand. Nobody can. The volume is real, and automated triage is a reasonable tool.
What is not reasonable is automation that only runs in one direction. A few changes would go a long way:
- Check the obvious defenses before acting. A permissive license in the repository is a fact a machine can read.
- Make repeat false claimants pay a cost, even a small one, so volume stops being free.
- Give developers a human contact when an automated decision removes their listing.
- Tell people which system made the call, so the error can be traced instead of guessed at.
Agents are getting better at doing things on our behalf. That is genuinely useful. But an agent with the power to remove software from the world’s largest app store, and no obligation to be careful, is not a productivity tool. It is a liability wearing a productivity tool’s clothes.
Luanti will probably be fine. Its community is loud and its license is unambiguous. The projects I worry about are the ones that get the same email and simply give up, because arguing with a machine is exhausting and nobody is listening on the other end.
🕒 Published: