Local stays local.
When you run a model privately on your own machine, the whole exchange stays there. No AI provider receives your prompt or its answer, and no cloud subprocessor is involved in that conversation.
Subprocessors
The companies that help us run Cuttlefish.
This is our current subprocessor register: the third-party companies we rely on to host, deliver, bill, and run Cuttlefish, what each one does, and the data it handles. Effective April 23, 2026.
What this page is
A plain, current list of who handles what.
A subprocessor is an outside company we use to provide part of Cuttlefish — for example, the servers that run it or the service that processes a payment. We keep this list current and give advance notice before a new or replacing subprocessor starts handling customer data, so you have a chance to review or object.
| Subprocessor | Purpose | Data categories | Processing region | Entity / jurisdiction |
|---|---|---|---|---|
| Microsoft Azure | Hosting and infrastructure — the servers, database, and file storage that run the Cuttlefish service. | Account and workspace data, content you ask Cuttlefish to store or process in the cloud, usage and diagnostic logs. | United States and other Azure regions in use, depending on the service. | Microsoft (United States). |
| Azure Front Door & CDN | Delivery and edge network for the Cuttlefish websites and app — serves pages quickly and helps keep them available. | Network request metadata such as IP address, browser type, and the pages requested. | Global edge network, served from the location nearest you. | Microsoft (United States). |
| Stripe | Billing and payments for paid and enterprise plans. | Billing contact details and payment information you enter at checkout; we do not store full card numbers ourselves. | United States and Stripe's global payment regions. | Stripe (United States). |
| LaunchDarkly | Feature delivery — lets us turn features on or off and roll out changes safely. | A non-identifying account or device identifier used to decide which features are active for you. | United States. | LaunchDarkly (United States). |
| AI model providers (your choice) | Running your request through the AI model you connect — for example a cloud model from a provider you choose. This only applies when you send a request to a cloud model. Models you run privately on your own machine stay on your machine and use no outside company. | The content of the request you choose to send to that model, and the response. | Varies by the provider you connect. | Varies by the provider you connect. |
Less than you might expect
Many users trigger few of these, or none.
Because you choose your model and can keep work on your own machine, a lot of what you do never reaches an outside company in the first place.
Only what's needed.
Each company above sees only the data it needs to do its part — billing details go to the payment processor, request metadata to the delivery network, and so on. We don't pool your data across them.
You hear about changes.
Before a new company starts handling customer data, or one is replaced, we update this register and give notice so you can review the change and raise an objection if you have one.
Diligence & contracts
Need the signed register for procurement?
For a countersigned subprocessor list or contractual data-processing terms, request the packet and we'll route it through legal review.