Fair Use Policy
Last updated: 11/07/2026
How we bill
AI Foundry no longer sells shared, flat-fee inference subscriptions. Models as a Service, hardware, and in-region deployments are priced per engagement — monthly or term commitments agreed in your order. Any usage-based metrics that apply are set out in your agreement; you will never receive a surprise per-token invoice.
Capacity and throughput
Because Models as a Service runs on a GPU dedicated to you, throughput is bounded by your own hardware — not an artificial per-key request cap shared with other customers.
Any limits specific to your deployment (for example, concurrency or rate ceilings for a given serving configuration) are set out in your agreement. Standard OpenAI conventions apply — requests beyond a configured ceiling return HTTP 429with a Retry-After header.
Acceptable use
You agree not to use the service to:
- Generate content that is illegal under New Zealand law.
- Operate at sustained volumes designed to circumvent concurrency limits (e.g. distributing requests across many low-effort accounts).
- Resell raw API access to third parties.
- Conduct security testing without prior written agreement.
Enforcement
We monitor for abuse at the aggregate level (traffic and account anomalies). We do not log prompt or completion content for enforcement. If we believe an account is violating this policy we will contact the billing email before taking action.