Data protection
Process discovery data protection
You choose where the files live. A cloud engagement runs isolated and encrypted. On the appliance, the corpus stays on your machine and ThinkPrior never receives it.
Cloud engagement
- Action StepIsolated environment
- Source · Not used to train modelsEncrypted
Appliance
- Action StepFiles on your machine
- Boundary · Stay on the machineYour keys
A cloud engagement is isolated. The appliance keeps the corpus local.
01 · A cloud engagement is its own environment
How is a cloud engagement isolated?
Each cloud engagement runs in an isolated environment hosted in Google Cloud's United States region.
Client data is encrypted at rest and in transit. It is not shared across accounts, and it is not used to train models. You can request deletion of engagement data at any time, subject to your agreement and the retention needed for security or legal obligations.
A cloud trial is provisioned for one engagement at a time while ThinkPrior is in user testing. Isolation is per engagement, so one account's document pile is not available to another.
US region
Hosted in Google Cloud's United States region.
Per engagement
One account's pile is not available to another.
02 · The appliance does not send the pile
Can ThinkPrior run without receiving the documents?
Yes.
The on-premise appliance keeps engagement documents on the machine you run. ThinkPrior does not receive those files. The appliance does not call ThinkPrior or Google Cloud for that corpus.
You operate the machine, and the maps and audit logs stay on it as well. Choose this path when a client NDA or an internal rule forbids uploading the pile to a vendor cloud. The discovery workflow on the appliance is the same document-to-map loop.
No corpus
The appliance does not send files to ThinkPrior.
Local logs
Maps and audit logs stay on the machine you run.
03 · You hold the keys you bring
Who holds the model keys?
You hold the model keys when you bring your own.
On-premise, keys stay on your machine across five self-serve slots: fast, reasoning, embeddings, vision, and speech. Routing does not log keys, prompts, or documents. Cloud enterprise points model calls at an OpenAI-compatible endpoint when access is provisioned.
That is not a switch on the cloud trial. The appliance is air-gapped only if those model services never leave your network. Otherwise model calls go to the endpoints you named, and the documents themselves still do not go to ThinkPrior.
Your keys
On-premise keys stay on the machine you run.
Not a toggle
Cloud enterprise bring-your-own-key is provisioned.
04 · Review documents are available on request
What can a security review ask for?
A security or procurement review can ask for the documents it needs before a pilot starts.
Enterprise customers can request the current subprocessor list, a data processing agreement, and answers to a security questionnaire through the contact on the About page. The privacy policy describes how cloud and on-premise data is handled, including the analytics on this marketing site. For a review that cannot accept any vendor cloud, the on-premise appliance takes ThinkPrior out of scope for the corpus, because the files never reach us.
On request
Subprocessor list, data processing agreement, questionnaire.
Out of scope
On the appliance, the corpus never reaches ThinkPrior.
Your data stays yours, in our cloud or on your machine.
Each cloud engagement is isolated. Or keep your documents on a machine you control.
Cloud engagements
Every cloud engagement runs in an isolated environment in Google Cloud. We encrypt client data at rest and in transit. We never share data across accounts. We never use data to train models. You delete engagement data at any time.
On-premise appliance
The appliance keeps your documents on your machine. ThinkPrior does not receive those files. The appliance does not call ThinkPrior or Google Cloud for those documents. You operate the machine.
Your models, your keys
Use your own keys. On-premise uses only the models you configure. Secrets stay on your machine. Routing never logs keys, prompts, or documents.
In our cloud, isolated and never used to train models. On your machine, never sent to us.