Your inference,
in your region.
Residency, retention and isolation are part of the deployment conversation from day one.
Data residency
We pin customer inference workloads to the jurisdiction you choose: the UK or the EU.
Residency terms go into your contract, and network controls enforce placement. Ask us for the providers and facility details behind your proposed deployment.
Nothing trained
We never train on your prompts or outputs, and we never send them to another company’s model API.
Content is processed in memory and is not written to logs or disk unless you choose to store files or enable a cache. Staff don’t look at it unless you ask us to, for example in a support case.
Dedicated means isolated
Dedicated Endpoints run single-tenant to the GPU. Private Nodes are physically dedicated servers.
Private Nodes have their own serving stack and cache, with private networking and customer-held encryption keys.
Documentation and evidence
Agree your processing and availability terms before your workload goes live.
Per-request audit logging and endpoint evidence packs are on the roadmap. Ask us for current availability and deployment documentation.