01 · Work
Start with an everyday task
Draft a general email, summarise permitted text or brainstorm an idea. Use the company’s approved model connection from one workspace.
Policy & evidence
Company rules, applied while the work happens.
The cloud workspace applies configured policy to outbound requests, then records the decision and the policy version behind it. Explore those controls here, and discuss the requirements for your Core deployment with us.
Routing concept
Cloud + local Core
Request from your team
“Summarise this client contract.”
Kubrius checks your policy
Cloud AI approved
Everyday, low-risk work
Kubrius Core local
Private execution on your premises
Core runs locally. Automatic cloud-to-Core routing is not yet connected.
Part of the work
Write a request. The controls travel with it.
Your team asks for help in the workspace. Kubrius checks the outgoing request and included conversation history before a configured model is called.
02 · Check
Apply company rules automatically
Supported identifiers are masked; detected restricted content is held. Staff receive an explanation when a request cannot continue.
03 · Understand
See the decision and the usage
Review the outcome, the policy version that applied and usage evidence. These controls cover requests made through Kubrius.
Detection uses a limited set of patterns. It can miss sensitive information, so permitted tasks and data must be agreed before a pilot.
Inside the workspace
Everyday help. Understandable decisions.
Illustrative policy examples, not generated answers or a live model connection.
“Draft a friendly invitation to our public webinar.”
Decision
Allowed by the example policy
Next
With a live connection configured, the request goes to the approved external model.
Your part
Review and edit the draft before using it.
“Rewrite this invitation for [email protected].”
Before call
The email address is replaced with a request-specific placeholder.
On return
Supported placeholders are restored in the answer.
Boundary
The remaining text goes to the model provider. Masking is not complete anonymisation.
“Summarise this confidential client contract.”
Decision
Held by the example policy
Next
The request is stopped before a model call. The workspace explains the restriction.
Boundary
It is not rerouted to Core. Automatic private routing is not connected in the cloud workspace.
Evidence you can understand
See what happened, and which rules applied.
Policy as code
Versionable rules
The current policy is stored as versionable JSON. Changes can be reviewed in Git and deployed through the configuration process. There is no built-in Git approval dashboard.
Decision history
The rules at the time
Stored snapshots link decisions to the policy and configuration that applied. Historical decisions remain understandable when policy changes.
Activity
Outcomes and usage
Review decision metadata, outcomes and usage records. This is operational evidence, not an immutable audit store or a compliance certification.
From written policy to a control draft
Make a policy easier to implement.
The existing Policy Mapper can help sketch how a policy clause might become a control. Its output needs human review; it does not automatically change the workspace policy.
Good questions
Clear expectations.
Does this cover personal ChatGPT accounts or every AI tool?
No. These controls apply to requests made through the Kubrius workspace. They do not monitor or block separate personal AI tools.
Can you detect every confidential detail?
No. Current controls recognise supported email addresses, UK mobile numbers and National Insurance numbers, plus some secret and restricted-content patterns. They can miss names, indirect identifiers and confidential context. Use only the data and tasks agreed for your pilot.
Does this guarantee compliance?
No. Kubrius provides request controls and evidence to support company oversight. The policy, data handling, operational setup and legal obligations still need appropriate review.
Start with one useful task
See how Kubrius could work for your team.
Arrange a walkthrough with Chris, or discuss a small cloud pilot. Bring an everyday task and the rules your business needs to apply. We’ll agree the scope, setup and usage before any pilot starts.
- A relaxed 15 minutes, directly with the founder
- See the workspace and discuss your first use case
- I'll reply personally. No spam, no drip funnel
Thank you.
Your message has been received. Chris will get back to you personally to arrange a time.