Security and data
Your data stays where you decide
HumaneCircle's trust layer is built for the review your security and risk teams run before any AI goes live. This page explains where it runs, how data moves, who can change what and what evidence you keep. For the full security pack, talk to our team.
Where it runs
In the environment you choose
Run it as SaaS, in your private cloud, in your own data center or fully air-gapped. In an on-premises or air-gapped deployment, your traffic, rules and evidence stay inside your environment. You keep the models and providers you already use.
SaaS
Managed by HumaneCircle, the fastest way to start a pilot.
Private cloud
In your own cloud account, next to the AI you already run.
On-premises
In your own data center, under your own operations.
Air-gapped
For environments with no outside connection at all.
How data moves
Checked before it reaches a model
One trusted path
Requests from your applications and agents, their answers and their tool calls go through the trust layer. Only an approved application and version gets through.
Sensitive data protected
Personal, financial and health data, card numbers and secrets are found and redacted before they reach a model or leave your systems. You can trace what was detected and what changed.
Provider keys out of your apps
Your applications call the trust layer, not the model provider, so provider keys are not spread across application code.
Agents stay in scope
Each request is scoped to one approved application and the user it acts for. Agents call only the tools and data that application is approved to use.
Who can change what
Every change has an owner and a record
Owners and access levels
Each AI application has an owner who grants access levels such as full access, edit, call or view. Workspace membership decides what each person can see.
Versioned rules
Every policy and rule change is versioned. A change needs a new reviewed version; nothing is edited in place.
Independent release approval
An authorized reviewer who did not create the version approves or rejects it, with a reason.
A stop switch
An authorized operator can stop an application at once. Stopping and resuming are recorded like every other decision.
Pogo, the assistant, proposes changes for a person to approve. It works within the access of the person using it and can never give itself more.
Evidence
What your reviewers can check
Every decision recorded
Each decision is kept with the application version, the rule and the source behind it.
Risky actions held
A risky action waits for a named approver, with the rule and its source shown in plain words, and the outcome is recorded.
Covered and missing
SafeAssure shows, for each release, the control that covers each requirement, the evidence and what is still missing, in a report signed by HumaneCircle, for your reviewers.
Your call
We give you controls and evidence. Your legal and risk teams make the compliance decision.
Questions
Common security questions
Can we see the security documentation?
Yes. Ask for the security pack through the contact form. It covers architecture, data handling and how we run security.
Do we have to send our data to HumaneCircle?
No. In a private cloud, on-premises or air-gapped deployment, the trust layer runs in your environment. Choose SaaS when it suits your data.
Do we have to change model provider?
No. Keep the models, providers, agent frameworks and tools you use today, such as Claude and Codex.
What does this website do with my data?
The website and Pogo, its guide, are separate from customer deployments. The privacy notice explains what is sent, what is kept and for how long.