What you gain and lose by running an AI model yourself
A model on your own computer is the only setup where your text never leaves the machine. It costs hardware, setup time and answer quality, because the models that fit on a normal computer are smaller than the ones behind hosted chats. A private hosted chat asks you to trust the operator's rules instead. For most personal questions that trade is fine; for data you are not allowed to send anywhere, only a local model or your own server will do.
"Private AI" is used for 2 very different things: a hosted chat that promises to keep little about you, and a private LLM that runs on hardware you control. They solve different problems, and choosing between them starts with what your text is.
On this page
What a local model gives you
When the model runs on your own computer, nothing you type is sent anywhere. There is no operator, no log on someone else's server and no policy that can change next year. It also works without an internet connection. For documents you are not permitted to share, such as client files under a contract, this is the one arrangement that needs no one else's promise.
What it costs you
Hardware first: a normal laptop runs small models, and the larger ones need a graphics card with a lot of memory. Setup second: you install the software, pick a model, and keep both up to date. Quality third: the models that fit on a home computer are smaller than those behind hosted chats, and on hard questions the difference shows.
What a private hosted chat gives you
A hosted chat runs a larger model with no install, on any device, in seconds. The price is trust: your text reaches the operator's server, and what happens to it depends on the operator's rules. Those rules can be checked. Does it need an account? Does it keep a log? Is the text used for training? Is it passed to another company? A chat that answers all 4 in writing gives you something to hold it to.
Private AI for a company
A business asking for a private AI platform is usually asking a contract question: where the data is processed, who can access it, and whether it is used to train anything. A model on the company's own server answers all 3 by design, and it also needs someone to run it. A hosted service answers them in its data processing terms, and those terms are what your legal team should read before anyone pastes in client data.
A simple way to decide
Ask whether you would be in trouble if this text were read by the operator. If not, a private hosted chat without an account is enough, and faster. If yes, run the model yourself or on your company's own server. If you are unsure, take the names, numbers and places out of the text first; most questions survive that, and then the choice matters less.
Questions about private AI
What is a private LLM?
A language model that runs on hardware you or your company control, so the text you give it is not sent to an outside service.
Is a local model uncensored?
Not automatically. It has no service filter in between, but its training still decides what it answers and how. Some open models are trained to refuse as much as hosted ones.
Where does this chat run?
On our own servers in the Netherlands, with no account and no profile. It is a hosted chat, so it suits questions where you accept that trade.
For everyday questions, ask here without an account or an install.
Open private chat