Terms of service

ArcemCloud AS, org.nr 933 755 630, Inkognitogata 8, 0258 Oslo, Norway ("ArcemCloud") provides three services on the ArcemCloud platform: Managed models (inference against a curated catalog of open-weight models behind an OpenAI-compatible API), Your models (deployment of customer-supplied model weights within the customer's own isolation boundary) and GPU compute (dedicated GPU instances with SSH access and per-second billing). The scope of each service is described on the platform pages and, for enterprise customers, in the applicable individual agreement.

Customers are responsible for the accuracy of their account information and for safeguarding credentials and API keys issued to their projects. The services may be used only for lawful purposes. In particular, customers must not:

• use the services in violation of applicable export-control or re-export rules;
• attempt to circumvent isolation boundaries, quotas, rate limits or metering; or
• use the services to infringe the rights of third parties.

Customers are solely responsible for the models, weights, data and workloads they upload to or run on the platform, including the legality of the training data behind any model they bring.

Service levels, support response times and availability commitments are defined in individual agreements. In the absence of an individual agreement, the services are provided with reasonable skill and care but without a committed SLA.

ArcemCloud commits, for all services and all customers, that:

• all customer data and workloads are processed in Norway;
• no US or other non-EEA processors are used;
• customer-supplied model weights remain within the customer's isolation boundary and are not accessed, copied or reused by ArcemCloud;
• prompt and completion content is not stored; and
• customers may export their data, weights and workloads and terminate at any time — there is no lock-in to proprietary formats or APIs.

Fees are stated in Norwegian kroner (NOK), exclusive of VAT, and metered as described on the pricing page: per-token for inference and per-GPU-second for compute and endpoint time. Quotas are hard limits; usage beyond a configured quota is rejected rather than billed. Invoices are payable per the terms stated on the invoice or in the individual agreement.

ArcemCloud may suspend or terminate access to the services in case of material breach of these terms, non-payment, or where required by law or by an order of a competent Norwegian authority. Where practicable, ArcemCloud will give notice and an opportunity to cure before suspension. On termination, customers may retrieve their data and weights within a reasonable wind-down period.

Neither party is liable for indirect or consequential loss. ArcemCloud's aggregate liability under these terms is capped at the fees paid by the customer in the twelve months preceding the event giving rise to the claim, except for liability that cannot be limited under mandatory law. Individual agreements may state different caps.

These terms are governed by Norwegian law. The parties agree to Oslo tingrett (Oslo District Court) as the exclusive legal venue.