The short version
A public chat window is the wrong place for client files, patient notes, or the week’s books. Private AI keeps that data on infrastructure you control.
OTR Digital runs local models on servers we host in Miami. SOC 2 Type 1 and Type 2 cover that hosting. Training is how most clients start, on the same privacy model as the build. A person still signs off on anything that reaches a client.
Key takeaways
- Client data stays private when it never leaves the environment you chose.
- A local model runs on hardware you control, not a public cloud that may train on what you send.
- OTR hosts on Miami servers we run, with SOC 2 Type 1 and Type 2.
- Training is the on-ramp. Most clients start there before a custom system.
- A person still signs off on the last step.
How to keep information shared with AI safe
Data privacy in AI is making sure client and patient information does not train someone else’s model and does not leave the environment you chose.
That is the job, not a policy PDF. Encrypted automations, gated access, and a person who signs off before anything reaches a patient or a client. OTR hosts on infrastructure we control in Miami, built for HIPAA and SOC 2.
The team is not asked to paste the week into a public chat window. If the work has to leave the room, it leaves under terms you can name.
See how that model sits on the solutions we build, and why AI projects fail when the office is asked to change how it already works.
What a local model is
A local model is software that runs on servers you can point to. The prompt, the files, and the output stay in that room. A public cloud model is a product used by many offices. It may train on what you send unless the contract says otherwise.
OTR runs local models on Miami servers we control so data does not train a public cloud. When a client wants a frontier model, it is wired in under enterprise terms with the same protections. That call belongs to the client.
The model is not the product. The hosting is.
What SOC 2 Type 1 and Type 2 cover
SOC 2 is an audit of how a company handles customer data. Type 1 looks at the controls on a date. Type 2 looks at those controls over a period.
OTR’s SOC 2 Type 1 and Type 2 cover private client data, local models, and servers we run. Your team stays audit ready on infrastructure we host and monitor in Miami. Patient data shapes how we host and how we ship.
Compliance is the floor. The daily practice is simpler: the data stays here, and a person still signs off.
Why training comes first
Training is the first step for clients integrating AI. It leads into the custom system: we build the workflow, connect it to the tools you use, and stay until your team is running it.
Training uses the same privacy model as the build. Local models. Servers we host. SOC 2 Type 1 and Type 2. The team learns the habit in the room, not from a PDF, and not by sending client files to a public window.
Most clients start there. Then the custom system sits on the login they already open. A person stays on the last step.
Photo: Taylor Vick, Unsplash.
Common questions
How do I keep my information I share with AI safe?
Keep it off public chat windows. Private AI runs on infrastructure you control, so client data does not train someone else's model. OTR runs local models on servers we host in Miami. SOC 2 Type 1 and Type 2 cover that hosting. A person still signs off before anything reaches a client.
What is a local model?
A local model runs on hardware you control, not a public cloud that may train on what you send. OTR runs local models on Miami servers we host. Client data stays there. When a client wants a frontier model, it is wired in under enterprise terms with the same protections.
What is SOC 2 Type 1 and Type 2?
They are audits of how a company protects customer data. Type 1 looks at the controls. Type 2 looks at those controls over time. OTR's SOC 2 Type 1 and Type 2 cover private client data, local models, and servers we run in Miami.
Why start with AI training before a custom system?
Training is the on-ramp. Most OTR clients start there before a custom build. The team learns the workflow on the same privacy model as the system: local models, servers we host, and SOC 2 Type 1 and Type 2. Then the custom system sits on the tools they already use.
What is private AI?
Private AI keeps your data on infrastructure you control, instead of sending it to a public cloud model that may train on it. OTR runs local models on servers we host in Miami. Client data stays there. A person still signs off on anything that leaves the system.
Get the hours back without adding headcount.
Book a demo to see how a custom system would run on your operation, and how we handle implementation so your team uses it.
