Prompts and responses are not stored by us
The service is designed to process prompts and model responses transiently and keep them out of databases, application logs, proxy logs, error logs and backups. Launch verification includes successful, failed and interrupted requests. This preview is not a completed security audit.
What we retain
Account and request metadata supports billing, usage exports and abuse handling. Request records include user and key identifiers, model, input/cached/output token counts, cost, IP address, user agent, timestamp and status. Request metadata is retained for 180 days, then purged from active records.
Payment and accounting records follow their required retention periods. Encrypted backup copies can outlive active-record deletion until the backup retention window expires. Read the draft Privacy Policy.
Where inference happens
abliteration.ai processes the inference request as an upstream provider. We do not host or train the underlying models. Its published policies describe transient content processing and no training on API content. Review its data-handling policy for details.
We do not claim Australian-only processing, anonymity, a certification or an uptime SLA. A model endpoint is not a local-only tool.
Controls you can use
- Separate engagement keys with budgets, expiry dates and model allow-lists.
- Revocation for an exposed or completed engagement key.
- Account rate limits and output limits.
- Usage exports containing metadata rather than client prompts.
Payments and account access
Stripe processes card payments; NOWPayments processes crypto payments. We receive payment references and status metadata rather than full card details. Email verification is required. Crypto checkout does not remove the AUP or account controls.
Report a concern
Contact support@redline.example for a security or privacy concern, or use the abuse reporting guide. Include only enough metadata to identify the issue; do not send live secrets or confidential client content.