1. Inference Data
This section applies to all inference requests processed by OpenInfer Engine, including requests routed through third-party platforms such as OpenRouter.
Prompt and Completion Data
OpenInfer does not use prompts, completions, or customer data to train, fine-tune, or improve AI models.
Standard inference requests are processed in memory to fulfill each request. OpenInfer minimizes retention of customer content and retains it only where necessary to provide requested platform functionality.
Session History
Certain APIs require server-managed conversation state, including APIs compatible with the OpenAI Responses API. For these features, OpenInfer may store session history so conversations can continue across multiple requests.
Session data is stored solely to provide the requested functionality and is never used for model training.
Where technically supported, stored session data may be encrypted. For self-hosted deployments, customers control the storage infrastructure and retention policies.
KV Cache
OpenInfer may temporarily store KV-cache fragments and related metadata to improve inference performance and synchronize cache state across compute infrastructure.
KV-cache exists solely to improve serving performance. It is not used for analytics, advertising, or model training.
For self-hosted deployments, equivalent caching infrastructure may be deployed entirely within the customer's environment.
Operational Metadata
OpenInfer retains operational metadata necessary to provide and secure the Services, including:
- timestamps
- token counts
- model identifiers
- latency metrics
- error codes
- billing and usage records
Operational telemetry is designed to avoid retaining customer payloads except as described in this Policy.
Customer-Authorized Debugging
Payload capture is disabled by default.
If requested by a customer for troubleshooting purposes, payload capture may be temporarily enabled with explicit customer authorization and only for the period necessary to resolve the reported issue.
Retention
OpenInfer retains customer data only for the period necessary to provide the intended platform functionality.
Operational metadata is retained for up to ninety (90) days for billing, reliability, security monitoring, and abuse prevention before deletion or aggregation.
Retention periods for session history and temporary KV-cache depend on deployment configuration and operational requirements. For self-hosted deployments, customers control retention policies within their own infrastructure.
2. Self-Hosted Deployments
OpenInfer Engine may be deployed entirely within customer-managed infrastructure, including private datacenters, edge environments, and air-gapped facilities.
In these deployments, customer content remains under the customer's control and OpenInfer has no access unless separately authorized for support purposes.
3. Customer Account Information
We collect business contact information necessary to create accounts, manage billing, provide support, and comply with legal obligations.
4. Website Information
When visiting openinfer.io, we collect standard server logs, including IP address, browser information, pages visited, and timestamps for security and operational purposes.
Information voluntarily submitted through contact forms or early-access registrations is used solely to respond to inquiries or provide requested information.
5. Security
OpenInfer maintains administrative, technical, and organizational safeguards designed to protect customer information.
These include:
- encryption in transit
- encryption at rest where supported
- role-based access controls
- tenant isolation
- network segmentation
- administrative audit logging
OpenInfer maintains a SOC 2 security program covering Security, Availability, Processing Integrity, and Confidentiality. SOC 2 documentation is available under NDA.
6. Subprocessors
For OpenInfer-managed cloud deployments, OpenInfer utilizes third-party infrastructure providers, including Amazon Web Services, GitHub, and Google Workspace.
Subprocessors process customer information only to the extent necessary to provide cloud infrastructure, storage, networking, development, or business operations. They are not authorized to use customer data for AI model training.
7. Data Residency
Inference requests may be processed according to customer-selected geographic regions or deployment-specific residency requirements.
Available regions for OpenInfer-managed infrastructure are published through the OpenInfer Models API where applicable.
8. Your Rights
Depending on applicable law, individuals may have rights to access, correct, delete, or restrict processing of personal information.
Privacy inquiries may be directed to [email protected].
9. Changes
OpenInfer may update this Privacy Policy from time to time. Material changes will be reflected by the "Last Updated" date and, where appropriate, customers will be notified.
OpenInfer, Inc. · San Mateo, California · [email protected]