Every question you type into a cloud AI leaves your computer. What happens to it next isn't a conspiracy — it's written down, in their own documentation. Most people have just never read it.
On ChatGPT's consumer plans — Free, Plus, Go, Pro — your conversations are used to train the model unless you go and switch it off. It isn't opt-in. It's opt-out, and the default favours them.
Not "an algorithm." People. OpenAI states that authorized employees and specialized third-party contractors may access your conversations — for engineering support, to investigate abuse, and for legal compliance.
Whatever you typed at 2am sits on a server, attached to your email address, readable.
During the New York Times' litigation, a court ordered OpenAI to preserve user chat logs — including conversations people had already deleted. OpenAI publicly fought the order.
That's the part worth sitting with. Once your words are on someone else's computer, deletion is a policy, not a fact. A court, a subpoena, or a breach can undo it — and nobody has to ask you.
Consumers pay with data.
Businesses pay with money.
Look at who gets privacy. ChatGPT Team, Enterprise and the API are not trained on. The free and cheap consumer plans are. Privacy isn't a principle in that model — it's an upsell.
| Plan | Trains on your chats? |
|---|---|
| Free · Plus · Go · Pro | Yes, by default |
| Team · Enterprise | No |
| API | No |
Ask a cloud AI 20–50 questions and roughly half a litre of clean water is evaporated cooling the servers.
GPT-3's training drank about as much fresh water as building 370 cars. That's one model, one training run.
Data centres already burn ~1.5% of global electricity. The IEA expects that to roughly double this decade.
Sources: UC Riverside · IEA · water estimates vary by method and location — we break the numbers down honestly here.
GreenCube runs the AI on your computer. Not in a warehouse. Not on someone's server. On the machine already in front of you.