Regal Technology Partners, a Tier 2 defense electronics manufacturer, runs its AI platform on its own data, inside its own secure environment. This is a plain explanation of how that platform uses large language models without ever exposing, sharing, or training on Regal's proprietary data.
The most important thing to understand: the large language model (such as GPT-6 Astra) is a pre-built tool. It was trained by OpenAI on public data, such as books, websites, and research papers, long before it ever reached Regal's environment. nBrain never changes the model. Regal's data is never fed to it to learn from. The platform uses it the way you'd use a calculator: give it a problem, it gives an answer, and it forgets everything immediately.
Imagine hiring a highly skilled engineer who went to the best schools and has decades of general knowledge. You bring them into your secure facility, hand them a specific document, and ask them a question about it. They read the document, give you their expert answer, and then you shred the document and wipe their memory of it.
That's how Regal's platform works. The AI has general knowledge from its training (school). Regal's documents are handed to it one question at a time (the specific task). It answers and immediately forgets. It never takes the documents home. It never learns from them. It never gets smarter from Regal's data.
Here's what happens when someone on Regal's team asks the platform a question. Follow the arrows.
Regal's documents and data, in Regal's own Azure Government subscription. Everything lives here.
Encrypted connection. Stateless processing. Not used for training.
Reads the question. Generates an answer. Immediately forgets. No learning occurs.
The arrows are an API call, the same way your phone talks to a weather service. A question goes out, an answer comes back, and nothing is kept on the other end. Microsoft's Azure OpenAI Service terms state that inputs and outputs are not used to train, improve, or modify the model.
There are two completely separate things happening. On the left is everything nBrain custom-built for Regal: this is where the intelligence specific to Regal's operations lives. On the right is the general AI model, which never changes.
This is where all the "training" happens, but it's not training the AI model. It's building the instructions, rules, and logic that wrap around the model.
The model's weights, the actual "brain" of the AI, stay exactly as they were shipped. Regal's data never touches them.
Here's the path from the moment someone types a question to the moment they get an answer.
"What are the thermal specifications for the connector assembly on this program?"Happens inside Regal's secure environment
The custom-built retrieval system searches Regal's document library (its vector database) and finds the three to five most relevant sections. The documents never leave Regal's environment: the search runs entirely inside Regal's Azure subscription.
Regal's data stays in Regal's environmentThe question and the retrieved snippets are packaged together and sent to the Azure OpenAI API over an encrypted connection. Think of a sealed envelope: only the API can open it, and it's processed in an isolated session.
Encrypted in transit, TLS 1.3The model reads the question and excerpts, writes a detailed answer, and returns it. Once the response is sent back, the model keeps nothing from the exchange. There is no session, no history, no learning. The next question starts completely fresh.
No data kept · No learning · No model changeThe engineer sees the answer in the platform, with citations pointing back to the original source documents so they can check it. The whole interaction is recorded in Regal's audit trail.
Answer and audit log stay in Regal's environmentThese aren't only nBrain's promises. They come from Microsoft's contractual commitments for the Azure OpenAI Service, in its data protection terms.
Prompts, inputs, outputs, and any data processed through the Azure OpenAI Service are not used to train, retrain, or improve Microsoft or OpenAI models.
With Azure OpenAI (as opposed to using OpenAI directly), Regal's data is not sent to OpenAI. It stays within Microsoft's Azure infrastructure. OpenAI has no access to the prompts or responses.
Each API call is processed in isolation. No other Azure customer can see, access, or benefit from Regal's data.
In Azure Government, Regal's data is processed in U.S. sovereign regions (US Gov Virginia and US Gov Arizona). It doesn't cross international boundaries or leave government-controlled infrastructure.
The word "training" causes confusion because it means two very different things in AI. Here's the distinction.
This is what people fear, and it's not what happens with Regal's platform. Model training means feeding data into the AI's neural network to change its weights and make it learn new information permanently.
nBrain writes instructions that tell a pre-built model how to behave for Regal's specific use cases. The model itself never changes; it just gets better directions.
Model training is like teaching someone a new language. It changes who they are. Once they learn it, they can't unlearn it, and they'll use that knowledge with everyone they meet.
Agent configuration (what nBrain does) is like handing someone a job description and a reference binder. They read the instructions before each task, do the work, and hand the binder back. They don't become a different person. They don't remember what was in the binder after the task. And the next person who works with them starts fresh with their own binder.
All source documents, embeddings, and search indexes live entirely within infrastructure Regal owns and controls.
The model weights are frozen. Regal's usage does not change, improve, or modify the AI in any way.
Each question is an independent API call. The model keeps nothing between calls: no session, no memory, no history.
The Azure OpenAI data protection terms prohibit Microsoft from using Regal's data for model training or sharing it with third parties.
Nothing is sent to OpenAI. Nothing goes to third-party servers. Nothing crosses borders. API calls stay within Azure Government infrastructure.
nBrain never creates derivative, fine-tuned, or trained models from Regal's proprietary information. The model is generic. The data is private.
nBrain builds the intelligence around the model: the agents, the logic, the workflows. Regal's data never goes inside the model. That is the architecture's core guarantee.
In one strategy session we walk through where your Brain would run, which models it would use, and exactly what leaves your environment and what never does.
Prefer email? hello@nbrain.ai