Where OpenAI fits in our work.

The first fact is that we build with OpenAI's models. Standard API access at OpenAI's published rates, billed to your own account, with every build's model choice written down where you can argue with it.

The second fact is that we are an OpenAI Select Partner, a tier in the OpenAI Partner Network. That recognition has its own page, including what it says about how we work and how to verify it in about a minute. This page is not about the badge. It is about the work.

For balance, the rest of the ledger: we hold no tier with Anthropic, Amazon or Microsoft, and we are not a reseller for anyone. Every build names its model, its fallback, and the reason for each choice, written down where you can argue with it.

A workflow is not one model. Different steps get different models: a small fast one classifies and extracts, a larger one drafts anything a person will read. Where an OpenAI model is the right tool for a step, it gets that step. Where it is not, it does not, and the routing table says which and why.

Each routed step has a second provider behind it, usually a different vendor, and we exercise the fallback in testing rather than trusting it. Models are retired on published schedules, so this is maintenance rather than paranoia. developers.openai.com ↗

The accounts are yours, not ours. API credentials are opened in your name and billed to you at OpenAI's published rates, so nothing of yours runs through a shared Westringia account, and nothing of yours sits next to another client's work. When we hand over, you change the passwords and we are out.

If your work needs the request processed inside Australia, OpenAI's own API will not do it. OpenAI offers regional storage in Australia, but regional processing is a separate control, supported in the United States and Europe with limited processing in the United Arab Emirates for specific models. Against Australia, the processing column reads none, and the documentation says OpenAI may process and temporarily store content outside the region to deliver the services. OpenAI docs ↗

The Azure route narrows rather than removes the problem. On pay per token, the newest OpenAI model Microsoft will process inside Australia East is GPT-4o dated 20 November 2024, plus GPT-4.1-mini. A GPT-5 family model inside an Australian region needs committed provisioned capacity, and the pay-per-token alternative is a data zone Microsoft defines as any Asia Pacific nation.

So where processing must stay in Australia, the route we can verify today runs on Amazon Bedrock rather than OpenAI. We re-check the model-by-model position monthly, and it said the same thing before we held the partner tier. Checked 8 August 2026.

Two provider facts do most of the work here. OpenAI states that data sent to its API is not used to train its models unless you opt in, and has said so since 1 March 2023. And retention is a separate question from training: OpenAI keeps API logs for up to 30 days by default for abuse monitoring. developers.openai.com ↗

Those are contract terms. They can change, we read them again when they do, and the date at the bottom of this sentence is when we last did. Last checked 8 August 2026.

The rest of the data story is ours rather than OpenAI's, and it lives on the how we work page: per-client credentials, what a zero-retention arrangement does and does not cover, and the Australian Privacy Principle 8 position when information leaves the country.

Every model and tool in this ecosystem has a shelf life, and OpenAI publishes its dates. developers.openai.com ↗ Three already on the calendar: fine-tuning closed to organisations that had never run a job on 7 May 2026, the evals platform is scheduled to shut down on 30 November 2026, and Agent Builder shuts the same day, with the documented migration path being to rebuild in code.

The difference between a planned migration and an outage is usually how early somebody knew. We track these dates for every client we run software for.

We take no payment from any vendor for putting their model in your business. Our prices are the fixed ones on the pricing page, and your OpenAI usage is billed to your own account at OpenAI's published rates.

That is the arrangement that lets section 03 exist. A firm paid per token of a particular vendor's model could not publish a page saying when not to use that vendor.