Your models, our platform.
You own what’s built from your data. While we work together, you license our domain models with every update and run them where you choose. If you leave, you take what’s yours, and you can buy out the version you run. We keep our platform, including the method we use to build models, the way any software vendor keeps its product.
Who holds what
- Your data, records and photosYoursOwned
- Labels your people create in the workflow: accepts, edits, rejections, correctionsYoursOwned
- Labeled sets and evaluation sets built from your dataYoursOwned. Returned or deleted at the end of the term, your choice
- Adapters trained on your dataYoursOwned
- Your configuration: taxonomy, reference tables, thresholdsYoursOwned
- Philotic domain models, trained on public, commercially licensed dataPhiloticLicensed to you for the term, with every update
- Inference code and container images that run your modelsPhiloticLicensed to you for the term
- Documentation and runbooks for deploying and running your modelsPhiloticDelivered to you
- The training pipeline, the method behind it, and our infrastructurePhiloticNever included. It is our product; retraining is part of the service
“Your adapter on our domain model” describes two distinct files with two distinct owners.
A model adapted to your operation is a separate weights file that sits on top of the model beneath it, so the split matches how the files work. During the term, your license to the domain model and inference code covers running them on our hosting, in your own cloud or on your own servers, and every update comes with it.
What we commit to on your data.
- Your models onlyNothing from your operation trains or evaluates a general Philotic model or another customer’s model. That covers every signal, not only images: decisions, edits, added items, and the accuracy metrics computed from them.
- No publishingWe don’t publish results specific to you without your written approval.
- Where it runsModels adapted to your operation run only where you choose, and only under a written agreement.
- Third-party dataData a third party controls, such as a partner, a carrier or a regulator, stays out of analysis and training until you confirm you have the right to use it, enforced on the server rather than by policy alone.
What you decide.
- FoundationWhichever open-weight foundation you are comfortable with. We test candidates on your task and share the results; the same approval gate applies to any of them.
- Where models runPhilotic-hosted and dedicated to you, in your cloud, or on your own servers.
- When AI turns onOff until a model meets an accuracy bar you approve, on held-out examples.
- ExclusivityAvailable in your field for a defined term. It never covers the platform, or models trained only on public data.
What you have, and what you keep.
What you have during the term, and what you take with you if you leave.
During the term
Every update, where you choose.
- You license the Philotic domain model with every update, and the inference code that runs it.
- Run it on our hosting, in your own cloud, or on your own servers.
If you leave
You take what’s yours.
- You keep your data, labels, labeled sets, evaluation sets, configuration and adapters.
- Retrain on a foundation model of your choice with everything you keep, or buy out the frozen version you are running.
- There is no free license that continues after you leave.
- A bought-out versionFor your internal use only. It is not transferable, comes with no updates, and does not allow copying the model or distilling it into another. It runs without our platform.
- Never includedThe platform, the training pipeline and our infrastructure stay ours.
Agreed now, settled together.
These are commitments we can agree to now, in an addendum to the master agreement. The specifics, such as retention windows, the accuracy bar, the hosting setup and the terms of a buyout, get worked out with your team during the pilot, once we have seen your workflow and data.
Start with the terms, then the model.
We’ll walk your legal and security teams through these commitments before any of your data moves.