Who Owns Your Robot Training Data? | Elacity
Gig workers worldwide are filming their kitchens to train humanoid robots, then selling the footage once, forever. Elacity offers a different deal: let the robot learn from data you never hand over.
The Robot Training Data Is Your Living Room. You Don't Own the Archive.
Somewhere right now, a person has a phone strapped to their forehead, folding laundry in their bedroom so a humanoid robot can learn to fold yours. They were paid a flat fee, clicked through a terms-of-service box, and handed over the most detailed footage of their private life they will ever record. When the training run ends, the video does not come home. It joins a corporate archive of robot training data they will never see again.
That is the deal on offer for the data your own life produces. Film your kitchen, your medicine cabinet, the drawings on your fridge, and sell the tape once, forever. You are not a supplier in this market. You are the raw material.
This is not a thought experiment. This spring, MIT Technology Review documented a booming gig economy of at-home data recorders: thousands of workers across more than 50 countries filming themselves cooking, cleaning, and doing chores so companies can sell the footage to the firms building humanoid robots. The recordings capture family photographs, personal belongings, medication, the layout of every room. Nobody filming knows how long the buyer keeps the video, who it gets shared with, or what happens to their living room once the model is trained.
The Failure of Selling Robot Training Data Once
The problem is not that this work exists. For many people it is real income where good options are scarce, and there is dignity in that. The problem is the shape of the transaction.
A one-time sale of a copy is the worst possible structure for data. Data is not a widget. It can be duplicated at no cost, poured into a thousand training runs, merged with other archives, and sold onward, all without the person who produced it ever knowing or agreeing again. The moment the raw file leaves your hands in the clear, every future use becomes someone else's decision. You did not sell one video. You signed away every copy of it, in every context, forever, for a single payment.
The file does not even have to sit still to leak. Security researchers documented that popular Unitree humanoid robots quietly streamed audio, video, and sensor data to remote servers during normal operation, without telling the owner. A machine standing in someone's home had standing permission to watch and report, and no one had to approve a thing. Same failure, worn two ways: a system that takes your data by default and asks nothing.
The Paradigm Shift: Data You Keep, Rights You Set
Elacity exists to change the shape of that transaction. The premise is plain. You should be able to let a robot learn from your data without ever handing the data over. Not a better price for the same bad deal. A different deal, where the file stays yours and the buyer pays to use it under terms you wrote.
This runs on Elacity dDRM, decentralised digital rights management, on top of ElastOS, an open-source runtime built so your own machine is the source of truth and the cloud is a guest. Your data becomes a Wealth Capsule: an encrypted, programmable good with rights and royalties written into it, that others pay to use and that you never have to surrender in order to sell.
1. The data is used, never handed over
Selling data today means shipping the raw file. With Elacity dDRM, your footage stays encrypted everywhere except one sealed moment inside a locked sandbox, where a training pipeline consumes it and the working copy is then wiped. The buyer gets what they actually need, a model that learned from your data. They never get the archive itself. It is the principle we lay out in our explainer on decentralised DRM: no device, and no buyer, gets to keep the key.
2. The key is used, never owned
The key that unlocks your capsule is never held by any single party, Elacity included. It is split across an owned quorum of independent machines, and each one re-checks your on-chain rights before releasing its share. The key is assembled for a fraction of a second to authorise one specific use, welded to that transaction, then wiped. Revoke the rights and the next training run simply fails to open the file. You are not trusting a promise to delete your data later. The architecture never lets anyone hold it in the first place.
3. The terms are yours, and they can carry royalties
Because the capsule is programmable, the rules travel with it. Use for a single training run, not perpetual ownership. This model, not that one. A royalty attached to each use, enforced at the moment the key is granted rather than promised in a contract no one reads. You are not selling a copy and hoping for the best. You are licensing an asset you still hold, on conditions that cannot be quietly rewritten after the sale.
Here is the honest edge. The cryptographic core, keys used but never owned, content sealed except at the instant of use, rights re-checked on every use, is built and running today. The friendly consumer flow where anyone wraps their own footage and lists it in a few taps is what Elacity is building toward, not something we will pretend is finished. We would rather show you the line than hide it.
The deeper shift is about who the value belongs to. A hundred thousand hours of footage from homes around the world is an asset. Right now it is someone else's asset, assembled from lives that were paid once and written out of the story. It does not have to be built that way. As we argued when AI licensing began paying catalog owners instead of the people who made the work, a receipt is not a stake, and tokenising the wrong wealth misses the assets ordinary people actually make.
Sasha Mitchell, Elacity's founder, puts it in one line: "The people who create the value should own it. That is the entire reason Elacity exists." The person folding laundry for the camera is creating the value. The archive should answer to them.
For twenty years the internet ran on one quiet arrangement: you generated the data, someone else owned the asset. Your home does not have to be the next thing fed into that machine for free. You were the product. Now you own the asset class. See what a computer you actually own looks like: Get ElastOS.