Security

Your data stays in your systems. We keep none of it.

The agent works on one dedicated workstation inside your environment, under its own login, and only two encrypted trips leave it: the AI read and our support session.

The data path

One store, two trips.

One workstation holds everything, two encrypted trips leave it and keep nothing, and one gold route carries each exception to your team with the reason.

At a glance

Nine facts your IT can check.

Data at rest

Only on your workstation: the agent, your rules, the run records.

The only store

Zero retention inference

The AI read runs on AWS Bedrock and keeps nothing; a BAA covers it where health information is involved.

Prompts and outputs never stored

Model provider access

None. Not used to train models, not shared with the model’s maker.

Retained: none

Transport encryption

Encrypted with TLS. Support sessions encrypted end to end.

TLS 1.2 or later

Outbound only connectivity

One workstation, outbound only, no ports opened.

No inbound rule · No tunnel

Named access

Named logins with two factor. Every session logged.

Two factor routing · Session log

Dedicated service identity

Its own user in each system, so every action traces to it.

Audit trail per action

OS level credential vaulting

In the operating system’s protected store: Windows DPAPI or macOS Keychain.

Never in a file, an email or a chat

Offboarding

Uninstall one app and we are gone.

One page offboarding note before go live

Questions your IT will ask

The longer answers.

How do you reach our systems?
Through one dedicated workstation, with Splashtop Business: the same category of remote access software IT providers and small businesses already use. A small app runs on that workstation; we connect from our own computer with our own named login and two factor. Nothing for your team to install beyond that app.
Do we need to open ports or change our firewall?
No. The app only makes outbound connections, like a browser. At a glance: Network
Can you see our network?
No. We can reach that one workstation, nothing else. There is no network tunnel into your office.
Who can log in, and how?
Named people only, with two factor on our account. Every session is logged on our side: who, which computer, when. At a glance: People
Is it logged?
Yes. Splashtop logs every session. The agent keeps its own action log on the workstation, and it works under its own login in your software, so every action it takes is attributable to it and only to it.
Can Splashtop, the company, see our data?
No. The screen stream and keystrokes are encrypted end to end; Splashtop transmits them but does not store them. In their own words, Splashtop “does not process, store, or have access to any of the users’ computer data.”
Do you have a BAA with Splashtop?
Splashtop states it never accesses the data, which is the same position your internet provider is in. Where health information is involved, our business associate agreements are with the parties that do touch data: with you, and with AWS, where the AI read runs under one.
Our IT wants to review it.
Good. We send a one page install note, this page, and Splashtop’s own security documentation. Your IT can inspect the app and remove it whenever they like.
Where does the work happen?
The agent runs on your own workstation, inside your software. Two trips leave it: the AI read and our support session. Neither keeps a copy. The AI read runs on AWS Bedrock, which does not store the content or use it to train models, under a Business Associate Agreement where health information is involved. Run logs stay on your workstation, in your own systems.
What exactly leaves the workstation?
Only what the one decision needs: the document being read and the question about it. Not the record. Not the history. The call is logged (who, when), not the content.
What runs on the workstation?
The remote access app, a browser, and our agent, under its own local user. Passwords live in the operating system’s protected store (Windows DPAPI or macOS Keychain), never in a file, an email or a chat message.
Does the agent depend on someone watching?
No. It keeps running whether or not anyone is connected, and reports status and counts to our side over ordinary HTTPS. Status and counts only.
What happens when our software changes?
It stops. Sumvant checks each screen against the map we built for it; when an update changes a screen so it no longer matches, that item is set aside for your team rather than guessed at. We remap and revalidate before it runs unattended again.
And this website?
No trackers, no analytics scripts. What the site collects, and how to ask about it, is in the Privacy Policy.
How do we shut it off?
Uninstall the remote access app, or disable the agent’s login in your software. Either stops us immediately. We also give you a one page offboarding note before go live.
What happens to the work and its records?
Work we do for clients is governed by the written services agreement, and where health information is involved, a Business Associate Agreement, signed with each client. The run logs were on your workstation all along; they stay with you.
And what you hold from this site?
Ask, and we tell you, correct it or delete it. The Privacy Policy says how and how fast.

What we will not do

Six things we never do.

Guess.

Every action is read back from the screen before it counts.

Unproven click, no click · Stop and ask

Improvise around your rules.

The rulebook is yours, in plain sentences, and only your team changes it.

Versioned rules · Decision trace

Run alone too early.

The first runs are supervised, and the second workflow is earned by the first.

Supervised week · No big bang rollout

Open your network.

One workstation, a named login, two factor, every session logged.

No ports opened · No tunnel

Keep your data somewhere else.

Run logs stay on your workstation, in your own systems.

AI read on AWS · Not stored

Call ourselves certified.

There is no HIPAA certification, and we do not claim one.

BAA signed wherever health information is involved

Agreements

On paper, with every client.

A services agreement with every client.

Scope, access, support and offboarding, written down before the first run.

A BAA for health information.

A Business Associate Agreement, signed before any health information runs.

Ask the hard questions first.

Send this page to whoever has to sign off, and if something is missing, write to [email protected] and a person will answer.

Book a demo
Book a demo