Private document AI processed on VersePDF-controlled self-hosted infrastructure.
In the normal production private-AI path, uploaded document content and document-AI prompts stay within VersePDF-controlled self-hosted processing instead of being routed to third-party AI model APIs.
Self-hosted by default
Document-AI processing is designed to stay within VersePDF-controlled infrastructure.
Fail-closed external AI posture
External document-AI routing is not silently enabled by the presence of a credential.
Enterprise-ready boundary
Enterprise deployments can apply stricter egress, retention and integration policies.
Self-hosted capability groups
This public document describes capabilities and data-routing boundaries without publishing internal implementation or processing-component details.
Optional external-integration boundaries
Self-hosted document AI does not mean the entire service is isolated from every external integration; VersePDF is not an air-gapped service by default. External integrations are separate from normal document-AI inference and only receive the information needed for the feature the user or deployment enables. Self-hosting does not by itself establish a particular regulatory certification or guarantee that every optional integration is disabled.
Deployments can also inspect the public privacy-status response at /v1/privacy/ai for the configured document-AI routing boundary.
Provisioning is not document egress
Software or model packages may be downloaded during deployment, maintenance or provisioning. That is different from sending a customer's uploaded document or document-AI prompt to the package publisher. Runtime document-AI processing follows the self-hosted production boundary described above.
Enterprise deployment requirements
Organizations with stricter egress, retention, logging or integration requirements can discuss a deployment profile that disables optional external integrations and aligns the service with the organization's approved data-flow policy.