FAQ

Questions, answered.

What OraBrain is, how it deploys, how it stays governed and secure, and why a specialized local, on-premise model you own is the right choice for regulated decisioning.

About OraBrain
What is OraBrain, in one sentence?

OraBrain is a specialized decisioning LLM you own. It reasons over the signals your existing systems already produce and returns a clear, evidence-backed decision, deployed inside your perimeter.

Is OraBrain a chatbot?

No. OraBrain returns structured decisions, a verdict, the factor that drove it, the supporting evidence, and a recommended action, built to be consumed by a person or an API. It is a decisioning system, not a conversational assistant.

Does OraBrain replace my existing screening, monitoring, or analytics tools?

No. OraBrain is a decisioning layer that sits on top of the systems you already run. Those tools produce the signals; OraBrain reasons across them and returns a decision. You keep your existing stack, and OraBrain can take in as many signal sources as you have.

What industries is OraBrain for?

OraBrain is proven today in financial services, across banking, payments, and digital assets. The underlying platform is domain-agnostic and applies to any field where decisions are high-stakes, must be explained, and cannot be handed to a public AI service, including security operations, critical infrastructure, defense and government, and healthcare.

Deployment & Integration
How is OraBrain deployed?

Three ways, all single-tenant: fully air-gapped, on-premises in your data center, or as dedicated private infrastructure we operate for you. There is no re-platforming to move between them, and no client data is ever routed to an external model provider.

How long does integration take?

Because OraBrain reasons over facts you already produce, integration is largely a matter of pointing your existing signals at its API and consuming the structured decisions it returns. It is a day-one activity, not a quarter-long project. Most deployments move from connection to operation in weeks.

Does OraBrain call any external AI service?

No. The intelligence is a model you hold and run in your environment. There is no third-party inference dependency, which is what makes air-gapped and accredited deployments possible.

Trust, Security & Governance
How can I trust an AI to make regulatory decisions?

OraBrain does not make the binding decision, your officer does. It produces a recommendation with its reasoning and the evidence behind it. Where your policy requires it, nothing becomes a binding action until a person approves it. You set the policy, you can override any verdict, and every decision is logged and reproducible.

Is every decision auditable?

Yes. Inputs, outputs, model identity, and timing are logged for every case, so any decision can be reproduced and defended to QA, model risk, or a regulator. Specialized models are versioned so you can evaluate, compare, and roll back.

Who is accountable for a decision, OraBrain or us?

You are. OraBrain is decision support. Accountability stays with your institution and your named decision-maker of record, which is exactly how it should be for a regulated process.

Why a Local, On-Premise Model
Why use a specialized local, on-premise model instead of a frontier model?

A frontier model is a rented, general-purpose service optimized to be good at everything for everyone. Regulated decisioning needs the opposite: a model specialized to your domain and policy, that you own and can run where your data governance requires. A specialized local, on-premise model can match or beat a much larger general model on a narrow, well-defined task, while being cheaper to run, faster to respond, easier to audit, and deployable air-gapped, none of which a frontier API can offer.

Can a local, on-premise model really be as good as a frontier model?

For a narrow, high-stakes task, yes. Frontier models spread their capacity across every subject imaginable. A model specialized on your data, policies, terminology, and historical decisions concentrates its capacity where it matters and reasons like an expert in your field, rather than a generalist reaching for a plausible answer. The goal is not the biggest model; it is the best decision on your task.

Isn't sending data to a frontier API a problem for us?

For most regulated institutions, yes. A frontier API means your cases leave your perimeter and run on infrastructure you do not control, which is often a non-starter for AML, sanctions, and sensitive customer data. Owning the model removes that exposure entirely: the data and the reasoning both stay with you.

What about cost and speed at scale?

Running a frontier model on every alert is expensive and adds latency you do not control. A specialized local, on-premise model is far cheaper per decision and fast enough to sit in the transaction path, which matters when you are deciding thousands of cases a day rather than answering the occasional question.

Do you use frontier models in OraBrain?

No. OraBrain runs entirely on local, on-premise models you own. No frontier model is ever in the decision path, and no client data is sent to one. We hold frontier models in high regard, and when we assess OraBrain's reasoning we have had them act as independent judges, but that is an external evaluation, never part of the product or your production pipeline.

Still have a question?

Our team is happy to walk your risk, compliance, and security leaders through any of this in detail.

Talk to our team
Talk to us

AI agents will transact financially. Static security will fail. Prax Labs is ready.