Overview
Swif Browser Security, Swif's browser extension, does most of its work locally, in the browser. For a few features, it uses AI (a large language model, or LLM) to improve accuracy, cut manual admin work, or explain errors more clearly.
AI is used only where it helps, and only with the minimum data each task needs. Admins can turn each AI-powered feature on or off.
Privacy: Swif never sends passwords, cookies, authentication tokens, browsing history, screenshots, or full page content to AI. Each feature sends only the limited details it needs, such as a page's title or button labels.
Swif Browser Security was previously called the Swif browser extension.
How it works
Local first. The extension relies on local logic for speed and privacy.
AI as a fallback. AI is used only when local detection isn't enough, when a feature needs to understand context, or when it needs to write a readable message.
On demand. AI runs once per page load, or when an admin starts a task, never continuously.
Minimal data. Each request includes only the details the feature needs.
Features that use AI
Feature | What it does | What's sent | When it runs |
Sign-up detection | Detects sign-up and login pages, to block unapproved account creation | A light outline of the page, such as button labels and headings. Never form inputs, screenshots, or user data. | Once per page load, when sign-up blocking is on |
Login page detection | Finds login pages, for features such as credential autofill | The page title, URL, and labels | Only when local detection is unsure. Most login pages are detected locally, without any network call. |
PII detection and redaction | Finds personal data entered in forms on monitored apps, and hides it before it's stored, for example as | Handled by a self-hosted detection engine. AI may help with rare cases. | When forms are submitted on monitored apps |
Team table parsing | Reads team and member lists in SaaS apps, for bulk user provisioning and deprovisioning | The structured list on the page | Only when an admin starts a provisioning task |
Error message translation | Turns technical errors into clear messages for admins | The error text | Only when local pattern matching can't explain an error |
Features that don't use AI. Detecting which account is signed in to AI tools, for LLM Tracking, doesn't use AI. It reads some of each AI app's own cookies or site data in the browser, and never sends their values to Swif.
Admin controls
Admins can:
turn AI use on or off entirely
choose which features use AI
choose where automation is allowed
This lets you match your organization's security and compliance requirements.
Data handling and privacy
Swif Browser Security:
doesn't send passwords, cookie values, or authentication tokens to Swif or to AI
doesn't access browsing history
doesn't capture screenshots or full page content
sends only the details each feature needs
hides personal data before it's stored
Detection runs mainly in the browser. These controls support compliance with frameworks such as SOC 2, ISO 27001, and NIST.