A Robot Took Their App Down and Nobody Answered the Phone

On August 27, 2026, the Luanti team published a post explaining that their Android app had vanished from the Google Play Store.

The cause was a DMCA notice filed on behalf of Microsoft by a company called Tracer.AI, claiming Luanti infringes Minecraft's copyright. The notice cited one thing: US Reg. #TX 8-192-097, the copyright registration for Minecraft Java Edition 1.9. It did not name a single allegedly copied asset. Not one texture. Not one line of code.

Luanti is a voxel game-creation platform. It ships with no games and no game assets by default. The team published a screenshot of every texture the app contains, which fits comfortably on one screen.

Here is the part that should make you sit up. The same company filed the same kind of notice against Luanti in March 2023. Luanti filed a counter-notice and won. The app came back 46 days later.

Forty-six days. The DMCA says platforms must restore material between 10 and 14 business days after a counter-notice. Google took more than triple that.

Because the mechanism has nothing to do with games.

Strip the story down and it reads like this. A third-party vendor runs automated detection at volume. It generates a takedown request. A distribution platform honors the request within hours because honoring it is cheaper than evaluating it. Your product disappears. You file a counter-notice into a form. Then you wait, with no phone number to call and no human on the other end.

Tracer's own marketing describes the model plainly: AI agents, 85 percent faster takedowns, 100 percent more reviews month over month, 44 percent more takedowns month over month. Those are volume metrics. Nothing in that list is a precision metric.

When a system optimizes for throughput and the cost of a false positive lands entirely on the accused party, you get exactly what Luanti got. In February 2026, an indie game called Allumeria was pulled from Steam by a notice from the same vendor. Microsoft eventually dropped it, but only after the story got loud.

The Luanti team said the quiet part out loud: "If we were a small company relying on income from the app, this would be especially devastating."

That is you. That is most of the people reading this.

Do this exercise before you need it. Draw every path a stranger can take to become a paying customer, then mark the ones you do not own.

Most early-stage products end up with a shorter list than the founder expected. One app store. One ad account. One payment processor. One domain registrar. One email sending service. One search engine sending 70 percent of the traffic.

Each of those is a door that somebody else can close, sometimes by mistake, usually without warning, occasionally without a person involved at all.

The question is not whether you can eliminate the dependency. You cannot. Google Play reaches billions of devices and F-Droid does not. The question is what happens in the 48 hours after a door closes, and whether you have a second door that already exists rather than one you have to build under pressure.

Write down the answer for each. If the answer is "we would be dead," that is a real risk sitting on your risk register, not a hypothetical.

Speed matters more than being right, because being right takes weeks and your users leave in days.

First, capture everything. Screenshot the notice, the listing, the dashboard, the timestamps. If the notice was routed through the Lumen Database, pull the public record. You will need this later and the console may stop showing it.

Second, file the counter-notice the same day if the claim is baseless. Section 512(g) of the DMCA sets the clock running from receipt, so every hour you delay is an hour added to the outage.

Third, publish a status page and tell your users where else to get the product. Luanti pointed people to F-Droid and to direct APK downloads on their own site. That was possible because those channels already existed.

Fourth, email your list. Not the app store. Your list. If you have been putting off collecting emails because signups felt like friction, this is the week that decision gets expensive.

Fifth, go public if the claim is obviously wrong. The Allumeria takedown was reversed after coverage in the gaming press. Public pressure is a slow, unreliable tool, and it is also sometimes the only one that works.

A continuity plan is not a document. It is a set of things that already work.

Own the relationship. An email list and a domain you control are the only two assets in your business that no platform can revoke. Everything else is rented.

Keep a second distribution channel warm. Not planned. Warm. A direct download, a web version, a secondary store listing, a partner reselling for you. It should have shipped in the last 90 days and someone should have tested it.

Separate your identity from your channel. If your only login is "Sign in with Google" and your only support inbox is a Gmail address tied to the same account that got flagged, one enforcement action takes out three systems at once.

Document the recovery runbook in one page. Who files the counter-notice. Who writes the customer email. Who posts the status update. What the message says. Draft the message now, while you are calm.

If you are mapping dependencies and failure modes for the first time and a blank doc feels like too much, a planning tool like Foundra, a Notion database, or an old-fashioned spreadsheet all work fine. The tool matters far less than the fact that the list exists somewhere other than your head.

Partly, and with real limits.

A direct channel gives you a place to send people during an outage and keeps a fraction of your revenue outside the platform's control. That fraction is usually small. Most consumer products see the large majority of installs come from the store, and no amount of good intentions changes user behavior overnight.

What it really buys you is time. It converts an existential event into an expensive quarter.

There is also a harder trend to plan around. The Luanti post notes that both sideloading and alternative stores are under increasing pressure on Android. The escape hatches are narrowing while the enforcement automation is speeding up. Build the hatch anyway, and do not assume it will be there forever.

One more thing worth stealing from the Luanti post: they are exploring perceptual hashing to flag potentially infringing assets in their own community catalog, with a human making the final call every time. That is the correct pattern for anyone building moderation into a product. Machines narrow the pile. People make the decision.

Treating platform terms as boilerplate. You are going to be judged by rules you did not read, applied by systems that never saw your product. Read the enforcement and appeals sections. Those are the ones that matter when things go wrong.

Building the whole company on borrowed reach and calling it distribution. If your growth model is "we rank well in the store" or "we get cheap installs on one ad platform," you do not have distribution. You have a subsidy that can be withdrawn.

Assuming an appeal is a conversation. It is a form. The Luanti case shows that even statutory deadlines can slip past without consequence. Plan for the slow path and treat the fast path as a pleasant surprise.

Automated enforcement is now fast, cheap, and largely unaccountable. Luanti was removed twice by the same vendor with no evidence attached, and spent 46 days offline the first time despite a legal deadline of 10 to 14 business days.

Your email list and your domain are the only distribution assets you truly own. Everything else is rented from someone who can change the terms.

A second channel is only useful if it already works. Warm, not planned.

Draft your outage communications before the outage.

The cost of a false positive falls entirely on you. Price that into how much of your business you route through any single door.

Can a competitor get my app removed with a false DMCA notice?

It happens. Filing a knowingly false notice carries legal exposure under Section 512(f), but enforcement is rare and the practical burden falls on the accused. Platforms act on notices quickly because that is how they keep their safe harbor.

How long does a counter-notice usually take?

The statute says the provider should restore the material between 10 and 14 business days after receiving a valid counter-notice. Luanti's 2023 counter-notice took 46 days to produce a restored app. Plan for the longer number.

Should I hire a lawyer for a takedown?

For a clearly baseless notice, filing the counter-notice yourself is usually workable, and platform forms are designed for that. Get counsel if there is any chance the claim has merit, if you are being asked to accept jurisdiction language you do not understand, or if revenue loss is large enough to pursue.

Is a web version a real backup for a mobile app?

Often yes, and it is usually cheaper than a second native app. Install friction is the main reason people abandon apps in the first place, so a working web version serves as both a backup channel and a lower-friction front door.

How many distribution channels should an early-stage company have?

Two that work, and a written plan for a third. More than that tends to spread a small team too thin before product-market fit.

What is the single cheapest thing I can do this week?

Start collecting email addresses, and export the list somewhere you control. It takes an afternoon and it is the one asset that survives every kind of platform action.

1. Luanti removed from Google Play due to baseless AI copyright notice (Luanti Blog, Aug 27, 2026)
2. DMCA notice record, Lumen Database
3. 17 U.S. Code Section 512, Limitations on liability relating to material online
4. Microsoft DMCA takedown forces Minecraft-like Allumeria off Steam (PC Gamer)
5. Hacker News discussion, August 27 to 28, 2026
6. Luanti Android app on F-Droid