01The class it belongs to
Businesses now buy AI by the seat, at about $20 a month, and more and more of their use goes to the cheaper, faster models: the top models' share of that use fell from 53% in August 2026 to 45% by early September (Ramp AI Index, 9 Sep 2026). That is the class the private model we run on our own machines belongs to — and the difference is what it is allowed to see: because it runs on your machines, it can read the whole file.
02Why not an enterprise plan from a cloud vendor?
European data residency, which one large vendor introduced in February 2025 for new enterprise, education and developer accounts, decides where your data is stored at rest; it does not change who holds it. In a US copyright case a court ordered the same vendor, in May 2025, to preserve users' chat logs, deleted ones included, until the court ended the order in October 2025; enterprise, education and zero-retention developer customers were carved out, which shows that what a vendor keeps, and for how long, is set by courts and contracts, and can change. And where your own rules require the protected parts of a file to be hidden before it is sent out, our October test found that 12 of the 20 questions per file, 60 % of them, could not be answered by any model, ours included, which was the honest outcome: the rule, not the model, costs the answers.
A model inside your building has no second custodian, and it reads the whole file.
03What you can hold us to
- Every number has a source. If we can't show where a figure came from, we don't publish it.
- Research comes first. Our products grow out of our research. When the two pull in different directions, the research wins.
- We show our evidence. Our claims point to published research and to test runs you can download and check for yourself.
- We're upfront about limits. Our paper lists what it doesn't prove, the proof page does the same, and so will the report you get from us.
- We don't train on your data. Nothing you send is used to train an AI model.