Chrome extension · privately distributed to organisations · v1.7.0

FOR ENTERPRISE

Put your organisation’s rules in front of the send button.

Safe Privacy Mask Enterprise applies the masking rules an administrator distributes, inside the input fields your staff type into, in the moment before they send.

Detection and replacement both finish on the device. What someone types never reaches a server of ours. It is a separate codebase and a separate distribution from the free extension, and it is not published on the Chrome Web Store.

Managed policy
Org-forced rules
Audit log
GPO, MDM and Admin Console delivery
No server of ours

Talk to us about a deployment Request an evaluation build

Pricing depends on the number of seats and the scope of the rollout support, so we quote rather than list. We aim to reply to support within two business days.

THE PROBLEM

Reduce the paste accidents without stopping the AI work

Ban the tools and the work stops; leave it alone and customer names and credentials go to an outside model as typed. The third option is to make your own rules take effect in the moment before sending.

The policy went out. Whether it is followed, nobody knows

A document saying “do not paste customer names into an AI” changes nothing inside the actual input field. The only thing that enforces a policy is something standing in front of the send button.

Leave it to personal settings and every team differs

When each person registers their own rules, some have and some have not. A replacement your organisation decided on only reaches everyone when an administrator distributes it.

Surveillance-style controls meet resistance

Recording everything people type and shipping it to a central store reads as monitoring. This product records the decision only; the text itself is kept nowhere.

HOW IT IS DEPLOYED

Distribute it, let it act, keep the decision

A deployment rides on the Chrome Enterprise machinery you already have. There is no new management server to stand up.

An administrator prepares the policy JSON

Replacement rules, presets, target sites, audit and gateway settings go into one JSON object. The keys are defined in the managed schema.
Push it to the devices

GPO and the registry on Windows, a configuration profile on macOS, or the Google Admin Console. The CRX is distributed privately, and you can self-host it.
Text is replaced in the field, before it is sent

In the fields you targeted, matching strings are replaced. Only the pre-send pass runs synchronously, so it reliably applies.
The decision, and only the decision, is recorded

What happened on which host is written down. The text is not. You can also have events forwarded to a collector your organisation runs.

# Windows / GPO — where the policy is written
HKLM\SOFTWARE\Policies\Google\Chrome\3rdparty\extensions\<extId>\policy

{
  "masterEnabled": true,
  "enabledLock":   true,
  "rules":         [ { "find": "Contoso Ltd.", "replace": "[customer]" } ],
  "targetSites":   [ "chatgpt.com", "claude.ai", "gemini.google.com" ],
  "audit":         { "enabled": true, "sink": "siem",
                     "endpoint": "https://collector.example.co.jp/spm" }
}

This is the shape of a real distributed policy. The full key reference and per-platform examples come with the technical pack at rollout.

ADMIN CONTROL

What an administrator decides

Values set by managed policy win over a user’s own settings, and anything you lock cannot be changed by the person using it.

Org-forced rules

Distribute replacements for customer names, project names and internal system names. Users can neither edit nor delete them. Whether users may add rules of their own is also yours to set.

Per-site enforcement and exclusion

Name the domains where masking is always on, and the domains where the extension never runs. The mode for an individual domain can be overridden too.

Locking the enabled state and the settings

Force the extension on across the organisation so users cannot disable it, and make the settings page read-only if you want to.

Pinning presets and region

Fix which detection categories are active — email, card numbers, API keys, region-specific ID formats — and set the region detection should assume.

Enforcing the audit log

Turn the audit log on or off organisation-wide, regardless of the user toggle. Retention and the forwarding destination are set by policy too.

A second check before sending (optional)

Send the locally-masked text to a DLP or LLM endpoint your organisation runs for a second opinion. Whether an error lets the send through or stops it is your choice.

AUDIT

The decision is what gets recorded

The audit log is off by default. When it is on, what it records is the pre-send decision — not what was written.

Recorded

Timestamp and hostname (never the full URL)
The decision — passed, replaced, blocked, cancelled, warned
The type IDs of the presets that matched
Whether the optional second check was involved

Never recorded

The text the user typed
The text after replacement
Full URLs, paths, query strings, cookies
API keys, endpoints, or any other credential

The recorded fields are a fixed list, and anything else is dropped at the storage boundary. Because it copies only what is on the list rather than filtering out what is not, a call site that passes extra data does not get it stored.

Events live on the device, in a buffer holding the most recent 1,000. Even when you choose to forward them to your own collector or SIEM, the on-device record remains. Standing up that destination is yours; we never receive the events.

DATA FLOW

No server of ours is on the path

Detection and replacement both finish on the device. The only outbound request is the second check, and only when you have configured and enabled it.

Processed on the device

Detection and masking happen inside the browser. There is no mechanism that sends what was typed, what was found, or where someone browsed to us.

The second check goes to your endpoint

What is sent, when you enable it, is the text after local replacement. The destination is an endpoint you configure and operate; we are not in it.

No ads, analytics or telemetry

No usage measurement, no advertising trackers, and no error reports sent anywhere.

Check the free edition’s permissions, storage and traffic →

PROCUREMENT

We do not become a processor of your data

There is no route by which we receive what your staff type, so a review does not gain another party to assess.

There are no subprocessors

Detection and masking use no external API and no external server. When the second check is used, the destination is your own asset.

No cross-border transfer to assess

The product has no path by which typed text crosses a border. If you place the second check’s endpoint inside your own jurisdiction, that is where it ends.

Documentation you can put into a review

A technical pack covering the permissions requested, the data flow, the fields the audit log records and the known limits is provided during evaluation.

What we do not have

We hold no third-party certification (SOC 2, ISO 27001 or similar), and there is no vendor-side console. Management runs through Chrome policy distribution.

COMPARE

How it differs from the free extension

The free one is a tool each person adds and configures for themselves. Enterprise is a separate product, for an administrator distributing one set of rules to everyone.

Aspect	Safe Privacy Mask (free)	Safe Privacy Mask Enterprise
Who sets the rules	The person using it	The administrator (users can be locked out of editing)
How it is delivered	Each user adds it from the Chrome Web Store	A privately distributed CRX, force-installed via GPO, MDM or the Admin Console
Where it runs	Major AI chats, plus the fields you register yourself	Always on, across the domains named in the policy
Audit log	None	Yes — no message text recorded, forwarding to your own collector available
Second check before sending	None	Optional, to a DLP or LLM endpoint you operate
Terms and support	Free, as-is, one shared support form	Quoted, with rollout support and a two-business-day first-reply target

See the free Safe Privacy Mask →

LIMITS

What it does not do

Before you take this any further, here is what the product does not take on.

It does not guarantee that every piece of sensitive information is detected and replaced. Both misses and false positives happen.
It does not replace your DLP, your access controls or your approval processes. It works alongside them.
There is no vendor-side console and no reporting product. Settings are distributed through Chrome policy; collecting the audit events is your side.
There is no SSO and no user directory integration. The design has no concept of an account at all.
Chrome is the supported browser. Behaviour in other browsers has not been verified.
Whether anything is actually sent is always the user’s action. The product never decides that for them.

Start with an evaluation

We begin by handing over an evaluation build and the technical pack — permissions, data flow, audit fields. Tell us roughly how many people it is for and what rules you want distributed.