Trust & product status
We'd rather you know exactly where each capability stands. This page carries the nuance so the rest of the site doesn't have to.
What Privexa is designed to do
Privexa is designed to support governed AI use: it applies organisation-defined privacy policy to sensitive information before it reaches an AI model, preserves enough meaning for the task, restores authorised context inside the boundary, and records evidence of what left.
Product status
| Module / modality | Status |
|---|---|
| Privexa Secure AI | Controlled access |
| Privexa Documents & Knowledge | Controlled access |
| Privexa Image Privacy | Controlled access |
| Privexa Large Image Processing | Controlled access |
| Privexa WSI | Under validation |
| Privexa Scribe | Pilot |
| Privexa Governance | Controlled access |
| Privexa API / Gateway | Controlled access |
Privacy architecture
durable instructions → versioned resolution → immutable derived protection → audited egress → scoped reconstruction
Each organisation's data, policy and authority are explicitly scoped.
Approved providers and models are enforced on the server, not chosen by the client.
Protection is derived from versioned state. Reprocessing creates new authority.
Approval is bound to the exact protected representation. Stale approvals do not silently transfer.
Evidence of exactly which protected representation was authorised to leave the boundary.
When protection cannot be established, the workflow stops rather than sending raw data.
Original context is restored only for authorised users, inside the boundary.
Approved providers and models are set by the organisation and enforced server-side. Current engineering supports configurable access to OpenAI, Anthropic Claude and Google Gemini models, with private/self-hosted models as a deployment direction. Specific models are approved per organisation, not assumed.
Image and slide protection include reviewer workflows. Generated outputs that matter — clinical notes, protected slides — are designed for human review before release.
Whole-slide privacy is a developed capability under extended validation. Evaluations confirm format coverage, throughput and reviewer workflow against your slides.
Managed, private cloud, on-premises and restricted environments are scoped per customer. Not every mode is qualified at production scale for every module.
Each AI egress is recorded against the exact protected representation, model and workflow that was authorised.
Security testing is part of Privexa engineering. We discuss testing scope and results directly with prospective customers during evaluation.
Responsible claims