Privacy Policy

RouteCortex operates as an inference infrastructure layer. We serve open-weight AI models on behalf of upstream platforms (aggregators, developer tools, chat frontends). We do not operate a consumer-facing product and we do not maintain end-user accounts.

What we receive

When an aggregator routes an inference request to us, the request contains:

• Prompt content (input messages)
• Model identifier and inference parameters (temperature, top_p, max_tokens, etc.)
• Metadata added by the aggregator (rate-limiting IDs, request IDs)

We do not receive end-user identifiers, IP addresses, or account information from aggregators. All requests reach us through the aggregator’s edge, not directly from end users.

What we log

Zero prompt or completion logging. Prompt content and generated completions never touch persistent storage on our infrastructure. All inference runs in memory only.

We retain the following operational metrics (no content) for capacity planning and reliability:

• Aggregate request counts per model and per hour
• Aggregate token counts (input and output) per model and per hour
• Response latency histograms (time-to-first-token, throughput)
• Error rates and failure classifications
• GPU utilization and memory metrics

Data retention

Operational metrics: 90 days rolling window. Aggregated monthly totals retained indefinitely for billing reconciliation with upstream aggregators.

Third parties

Model inference runs on rented GPU compute. Prompt content passes through the compute environment executing the model and is not shared with any upstream infrastructure vendor beyond what is required to run the workload. Cloudflare handles TLS termination and DDoS protection at the edge; no prompt or completion content is retained by the edge tier.

We do not disclose the identity of individual compute vendors for operational and commercial reasons. Aggregators requiring specific infrastructure disclosures (data-residency, sovereign-cloud, or vendor attestation) can request them under a mutual NDA.

End-user rights

Because we do not maintain end-user accounts or records, we have no end-user data to correct, export, or delete. Requests of that nature should be directed to the aggregator (the platform through which the end user accessed our inference).

Contact

Written correspondence: contact@routecortex.io