Responsible data
Ariamena describes its practices plainly. We do not claim certifications we have not earned or make promises the work cannot keep. What follows is how programs are designed and run.
Six principles
Each principle is an operating rule, not a value statement. They shape how programs are scoped, staffed, run, reviewed, and delivered.
Privacy-aware by design
Programs are scoped to what the model needs and no more. Sensitive material is minimized, protected, and handled according to agreed rules.
Consent-conscious workflows
Contributors know what is being captured, why, and how it will be used, before capture begins.
Clear governance
Every program has defined owners, permitted uses, and handling standards, documented from the start.
Quality and traceability
Each data point can be traced to its origin, its review history, and its acceptance status.
Context before scale
A smaller dataset that represents reality is worth more than a larger one that flattens it.
Respect for people and places
The people and environments behind the data are partners in the work, not raw material.
In practice
What responsibility looks like at each point in a program.
Before capture
Scope is agreed and written down: what is captured, where, from whom, for what permitted use. Contributors are briefed and consent is recorded before any recording begins.
During capture
Sensitive material is minimized at the source. Contributors can pause or withdraw. Program owners are reachable throughout.
During structuring and review
Access is limited to the people doing the work. Every data point carries its origin and review history.
At delivery and after
Permitted uses travel with the data. Retention and deletion follow the agreed terms. Documentation is delivered with every version.
Security and data handling
No certifications are claimed. This is what we do, stated plainly, and what we are working toward.
Storage
Data is encrypted in transit and at rest. Programs can be pinned to a storage region, including in-region options for the Middle East and Europe.
Access
Access is limited to the named program team. Contributors see only their own sessions. Client reviewers get read-only access to their program.
Retention and deletion
Retention periods are agreed per program. Raw material is deleted after delivery unless the agreement says otherwise. Deletion is confirmed in writing.
Contributor protection
Faces, screens, and documents can be blurred at capture. Contributors can pause or withdraw a session, and withdrawn sessions are removed from every delivery.
Agreements
Every program runs under a data processing agreement and mutual NDA. Sub-processors are listed in the program documentation.
In progress
A SOC 2 Type I readiness assessment is under way. We will state the result here when it is done, not before.
Tell us what your AI needs to understand. We'll help shape the data program that gets it there.