Privacy
What Safe-Desk and its browser extension do with your data.
This covers the Safe-Desk website at desk.sophiaxt.com, the Gate API, the AI connector and the Safe-Desk extension for Chrome. SOPHIA XT LLC runs all four. It describes the system as it runs today. Effective 23 September 2026.
The short version
Safe-Desk exists to keep client data out of AI tools. Text you type, paste or send through Safe-Desk is checked on our server, and client details such as Social Security numbers, account numbers and names are swapped for placeholders like [SSN_1] before an AI model sees the text. Passwords and keys are stopped. We do not store the text you send, the AI's answers or the values we catch. Files you upload for the AI connector are the one exception, described below. We keep a short record that a check happened, who ran it and what kind of thing was caught. We do not sell data, use it for advertising, or use it to train AI models.
Who is responsible
Safe-Desk is sold to firms. Your firm decides who gets an account and which rules apply, and its owner can read the firm's activity log. SOPHIA XT LLC operates the service for the firm. Questions about how your firm uses Safe-Desk go to your firm's Safe-Desk owner; questions about this policy go to us at contactus@sophiaxt.com.
Text you check
When you use the desk, the Gate API or the extension, the text goes to our server over HTTPS. The server reads it against your firm's rules in memory, swaps what it finds for placeholders and sends back the safe version. It does not write the text to the database or to a log. Error records are scrubbed before they are kept: card numbers, Social Security numbers, passwords, keys, bank and routing numbers, EINs, medical record and health plan numbers, emails, phone numbers, dates of birth, and detected names and addresses are replaced with a placeholder.
The table that turns placeholders back into real values is sealed with AES-256-GCM and bound to your account before it leaves the server. Your browser holds that sealed copy; it cannot be read or changed there. When an answer comes back, the sealed table and the placeholder text are sent to our server, which opens the table for that one request, puts the real values back and returns the result. The table is not stored on our side.
Files for the AI connector
When you upload a file or paste text under Files, we check it once. If it holds passwords, keys or privileged material, it is refused and nothing is kept. Otherwise we keep the protected text (placeholders for what your rules protect; anything set to Flag stays as written) and the real values, both encrypted and tied to your account, for 7 days or until you delete them. The original file is never kept. Documents your AI finishes through the connector are kept encrypted for 7 days and only you can open them. Encrypted copies of files and documents in our daily backups are gone within 14 days after that. For each AI app you connect we keep its name, where it returns to, and when it was connected and last used, while the app is connected and for up to 30 days after it was last used.
The Chrome extension
The extension runs on chatgpt.com, chat.openai.com, claude.ai, gemini.google.com, copilot.microsoft.com, copilot.cloud.microsoft, m365.cloud.microsoft (Copilot Chat pages only), www.perplexity.ai and chat.deepseek.com, and talks to one server: desk.sophiaxt.com. It loads no code from anywhere else.
It sends our server the text you paste or send in the chat box, so it can be checked, and the placeholders in the AI's answers, so the real values can be shown to you in the extension's own panel. The chat page keeps the placeholders. With checks goes the name of the site (for example ChatGPT or Gemini) and the sealed placeholder table for that conversation; restore requests carry only the placeholders and the sealed table. You connect the extension either by signing in with Safe-Desk, in a Safe-Desk sign-in window that Chrome opens, or with a one-time code from the Safe-Desk website. Either way it sends a label for the browser, such as "Chrome on Windows".
Every 5 minutes, and when Chrome starts or you sign in or out, the extension tells Safe-Desk its version, the kind of device, which sites it covers and a random ID for this browser. When it stops a file, it reports the site, the file type and a rough size, never the name or contents. A stopped file is also recorded in your firm's activity log, marked as reported by the device, and your firm's owner is alerted.
In the browser, the extension keeps its connection key, your name, your firm's name and whether protection is on, in Chrome's local extension storage. It keeps the sealed placeholder tables in Chrome's session storage, which Chrome clears when the browser closes. Chat pages cannot read any of this. Signing out from the extension, or disconnecting the browser from the Safe-Desk website, removes the key.
The extension does not read your browsing history, does not run on other sites, and does not collect anything from pages beyond the chat box, file uploads and the AI's answers on the sites above. What reaches the AI site is what you send there, with client details already replaced; those services handle it under their own terms and your own account with them. The extension's use of information follows the Chrome Web Store User Data Policy, including its Limited Use requirements.
Accounts and sign-in
For each person with an account we store a name, a work email, a password hash (scrypt) and a role. Names and emails are encrypted in the database with AES-256-GCM, and sign-in finds an account by a keyed hash of the email rather than the email itself. Sign-in sessions last up to 12 hours; we store only a hash of each session key. Each connected browser has its own key, stored only as a hash, with its label and when it was connected and last used, and it can be disconnected from the website at any time. Your IP address is used in memory to limit repeated sign-in and pairing attempts. Apart from the network address of each signed-in computer, described below, our application does not store it, though our host's standard web server logs may.
One computer per person
Each staff member can be signed in on one computer at a time. The extension and the Windows app send a random computer id, made when they are installed. It identifies the install, not the person. We keep the network address of each signed-in computer for 30 days, to enforce one computer per person. The owner and IT admin see, for each person, how many distinct network addresses were used over the last 30 days and when each computer was last seen, and can sign a computer out. The owner gets an alert when someone tries to use a second computer. When an extension or Windows app sign-in is signed out or disconnected, its record is deleted 30 days after it was revoked.
Healthcare practices
Safe-Desk does not sign a Business Associate Agreement, so medical, dental and therapy practices must not submit patient information (protected health information). When a healthcare practice signs up, the owner confirms this. We keep the time and who confirmed it in the firm's audit trail.
What we keep, and for how long
| Data | Kept |
|---|---|
| Text you check in the desk, the Gate API or the extension, AI answers, caught values, placeholder tables | Never stored |
| The digits from your last few extension or Windows app sends, protect-and-restore checks or Gate API calls, so a number split across messages is still caught | In server memory only, never on disk or in a log, for up to 30 minutes |
| Uploaded files: protected text and sealed values | 7 days, or until you delete them. Encrypted copies in our daily backups are gone within 14 days after that. |
| Finished documents | 7 days. Encrypted copies in our daily backups are gone within 14 days after that. |
| Connected AI apps: name, return address, when connected and last used | While the app is connected, and up to 30 days after it was last used |
| Activity log: who ran a check, when, the outcome and which kinds of data were caught | 400 days |
| Audit trail: sign-ins, failed sign-ins, staff, setting and billing changes, and a healthcare practice owner's confirmation that staff will not submit patient information, with its time and who confirmed (encrypted) | 400 days |
| Error records, scrubbed of client data | 90 days after last seen |
| Your firm's client list, as one-way keyed hashes only | Until the owner replaces it |
| Topics your owner sets up (name and phrases, encrypted) | Until the owner removes them. A topic the owner offers to other firms is reviewed by SOPHIA XT and, if approved, its name and phrases are shared with firms in the same industry, without your firm's name |
| Password reset requests (who asked, when) | 400 days after they are answered |
| Usage counts: messages and AI tokens per month, and what hosted AI cost each month | While the account exists, for billing |
| Accounts | While the account exists |
| Connected browsers and computers | While signed in, and 30 days after sign-out |
| The network address of each signed-in computer, to enforce one computer per person | 30 days |
| Revoked device sign-in records (extension and Windows app keys that were signed out or disconnected) | Deleted 30 days after they were revoked |
| Device events (stopped or warned uploads and pastes, sign-outs) | 90 days |
| Browser push subscriptions for the owner's phone and browser alerts | Until the owner removes that phone or browser from alerts, or its push service says it is gone |
| Daily backups of the database, encrypted whole | 14 days |
| Card details | Never; Stripe holds them |
Who else processes data
We use a small number of service providers. None of them receives a value your firm's rules swap out or stop.
| Provider | Why | What it receives |
|---|---|---|
| Hostinger | Hosting the service | The application and its database |
| Stripe | Billing | The owner's name, email, billing address and card, the number of seats, and what the firm's hosted AI use cost, in cents. Never message text or a client value. |
| An AI routing service, and the model maker your firm picks: Anthropic, OpenAI or Inception | Writing answers in the chat on Safe-Desk with the AI inside Safe-Desk, once Claude and ChatGPT models are offered there (today Mercury answers, through Inception, below). The routing service passes placeholder text to the model your firm chooses | Message text and any files attached to it, with caught values already replaced by placeholders. Values your firm set to Flag are sent as written. Handled under SOPHIA XT's account |
| The AI provider your firm connects: Anthropic or OpenAI | Writing answers in the chat on Safe-Desk, only when the owner pastes the firm's own key | Message text and any files attached to it, with caught values already replaced by placeholders. Values your firm set to Flag are sent as written. The provider handles it under your firm's own account and terms |
| Inception (Mercury) | Only when Safe-Desk's own model is used: chat answers for a firm that has not connected a provider, and suggesting phrases when an owner asks for them on a topic | Message text and any attached files, or a topic name, with caught values already replaced by placeholders. Values your firm set to Flag are sent as written |
| ChatGPT or Claude, through the AI connector | Only when a staff member connects it | The placeholder version of files that person uploaded, their titles and the kinds of data replaced; never a value |
| The push service of the owner's browser maker (for example Google for Chrome, Apple for Safari) | Phone and browser alerts, if the owner turns them on | The alert line, encrypted so the push service cannot read it; never a value |
| ntfy.sh | Phone alerts, only if the firm's owner turns them on. Also our own alerts when someone asks for a password reset or the server records an error | The firm's name, the staff member's name and the kinds of data caught. Billing notices such as a failed payment or the monthly AI spending limit, on the owner's alert. For a password reset request, the person's role, with the firm's name on the owner's alert and only a firm number on ours. Our own error alerts carry the error message, scrubbed of client data, and the page it happened on. Never a value |
If you use the extension or the AI connector, ChatGPT (OpenAI) or Claude (Anthropic) receives what you send there or the placeholder version of your files, through your own account, after Safe-Desk has swapped out client details. We do not have access to those accounts.
What Safe-Desk cannot promise
The rules catch identifiers with a known format reliably. Names, addresses and dates without a list to match against are detected from context, which is a best effort. The extension protects text boxes and file uploads on the sites above while it is signed in and your firm's subscription is active. It cannot stop someone typing client data into another site, another browser or voice input. Real values shown back to you are on your screen and in anything you copy them into. Safe-Desk helps a firm meet its confidentiality duties; it does not make a firm compliant with any law on its own, and it has not been audited under SOC 2.
Your choices
You can sign out of the extension or remove it from Chrome at any time. You can see and disconnect your connected browsers in Settings on the Safe-Desk website. Your firm's owner can see the firm's activity log and audit trail. To ask for a copy of your account data, a correction, or deletion of your account or your firm's data, email contactus@sophiaxt.com. Deleted data leaves our backups as they expire, within 14 days.
Changes and contact
If this policy changes, we will update this page and its effective date, and tell firm owners by email before a change that affects what we collect. SOPHIA XT LLC, sophiaxt.com, contactus@sophiaxt.com. The security page has more on how the service is protected.