Skip to main content

Conditions and actions

Updated on 10 November 2026.

A browser control tells the Browser Extension what to watch for and what to do when it happens. This page explains each part of a control. For the controls list and the steps to add or edit one, see Browser controls.

You set these parts in the control form under Govern > Policy Engine > Endpoint Agent > Browser.

Go toQuilrAI ConsoleGovernPolicy EngineEndpoint AgentBrowser

Key terms​

TermMeaning
Use caseThe activity the control watches for, chosen in When this happens. It sets the posture (the risk area), the risk behavior (the kind of activity) and the mandatory conditions.
Mandatory conditionsConditions the use case adds.
Additional conditionsConditions you add to narrow the control.
ActionsWhat happens when the control matches, set in Perform action.
ModeMonitor records matches without interrupting the user. Enforce also runs the Just in Time and Activate Agent actions.
CriticalityHow serious a match is.

Choose what the control watches for​

Besides its mandatory conditions, the use case decides which attributes and actions you can choose. See the Use case reference.

  • The use case is fixed once the control exists. To use another, duplicate the control and choose the new use case in the copy.
  • Some use cases need an additional condition before you can save. The console shows: "Add more conditions to this control to prevent business disruption."
  • Some copy and paste use cases do not support spreadsheets, such as Excel or Google Sheets. The console warns you.
  • Do not create controls on AI application blocked by organisation or SaaS application blocked by organisation. They can stop blocked apps from being blocked. See System controls compete too.

Narrow the control with conditions​

A condition is one test on one fact, for example Smart Group is not in Finance.

  • Mandatory conditions carry the Required tag and cannot be removed. A few let you choose the value.
  • Additional conditions: each attribute can appear only once, except URL.

How conditions combine​

  • Mandatory conditions and your first additional condition are joined with AND.
  • Each later condition card starts with an AND / OR menu.
  • Add condition to group adds a row inside a card and makes it a Group condition. Each row after the first has its own AND / OR menu, which starts at AND. Groups cannot be nested.
  • A group's rows are read from top to bottom: "A OR B AND C" means "(A OR B) AND C".
Use OR only inside a group

Browsers ignore a control whose top-level condition cards are joined with OR, so it never matches. To match either of two conditions, put both in one group and set the group's second row to OR. Keep the menu between cards at AND. For several values of one attribute, use is in. See OR between top-level condition cards.

Which operators you can use​

The Operator menu shows only the operators that fit the attribute.

  • Text comparisons ignore letter case.
  • is not equal to also holds when the value is unknown.
  • is in holds when the value matches one of the values you select. For multi-value attributes such as Smart Group, one match is enough.
  • is not in holds when no value matches your selection. A user in no Smart Group always matches Smart Group is not in.
  • For two-value attributes, such as Authentication Type or Yes / No attributes, use is equal to or is not equal to.

What each attribute tests​

  • User, the person signed in to the extension: Department, Smart Group, Status (Active or Inactive), Risk Level, Title, Location, DisplayName and Has account on any AI app.
  • Application, the app on the current page. You set its status, criticality and blocking in App posture and access requests.
    • Category, Name, Domain, URL and Headquarters.
    • Status (Needs Review, Approved, Tolerated or Unapproved), Criticality (Critical, Not Critical or Not Set) and Block status.
    • Is newly introduced, IdP status (Yes when the app is an identity provider used for single sign-on) and Multi-Factor Authentication (MFA) support. Unknown MFA support counts as Yes.
    • AI Mode (for A user is visiting an application): Yes on Google AI Mode pages.
    • Chatbot (for the chatbot use cases): Yes when the user posts into a chatbot on the page.
    • Login User Domain (for A user signing in and signing up to an application): the domain of the email address typed into the form.
  • Browser, the activity: User Action and Authentication Type (Login or Sign Up).
  • Data, what QuilrAI found in the content: Sensitive and Model (the detection models that found data).
  • Upload, the file: File Size in bytes., File Extension (files inside an archive are checked too), Label Name (the sensitivity label), Classification Labels and Encrypted (password-protected).
  • Source Application, the other app:
    • In the paste use cases and A user is copying sensitive data, the app the data was copied from: Category, Name, Domain, Status, Criticality, IdP status, Is newly introduced, Multi-Factor Authentication (MFA) support and Sensitive Data. For data copied from a desktop application, use Copied From Process, the application's name.
    • In the password reuse use cases, the app whose password is reused.
  • Account Owner, for A user is using shared credentials: the directory record of the account used to sign in, not the user (Department, Smart Group, Status, Risk Level and Title).
Content checks need the Browser Utility

Conditions on copied, pasted or uploaded content (Sensitive, Model, Label Name, Encrypted) and Copied From Process need the Browser Utility, a desktop program installed with the extension. See Content checks need the Browser Utility.

Choose what happens when the control matches​

Which action types you can use​

Action typeRunsWhat it does
Just in TimeIn the browser, during the activityShows a popup or redacts data.
Remediation ActionQuilrAI backend, after the activity is recordedIts only action, Send Alert to Syslog, sends an alert about the match to Microsoft Sentinel installations that have the Alerts & notifications capability. Despite its name, it does not send to the Syslog integration.
Activate AgentQuilrAI backend, after the activity is recordedStarts a Quilly agent.
  • A control can have one Just in Time action, plus actions of other types.
  • The two blocked application use cases offer no Just in Time action.
  • When several controls match the same activity, the extension acts on only one of them. See Set how serious a match is.
  • A control with no Just in Time action never shows a popup. If the extension acts on it, the other matching controls show none either. To only record activity, use Monitor: Safari records nothing for an Enforce control without a Just in Time action. See Enforce controls without a Just in Time action.

What users see​

The first word of a Just in Time action's name tells you what the user sees.

Name starts withThe user seesThe activity
PromptA card with a message and a button such as Acknowledge.Goes ahead.
Justify or JustificationA dialog that asks for a business reason, picked or typed (20 to 140 characters).Goes ahead on Submit. Closing the dialog blocks it.
BlockA dialog that explains the block and, by default, offers Request Approval.Is stopped. A block on visiting an app also freezes the page. Other blocks show Go Back.
AllowThe sensitive data window.Waits for the user's choice.
RedactA Sensitive Data Redacted notice.Goes ahead, with each detected value replaced by X characters.

Request Approval sends the user's reason to Action requests. An approval stops that action's controls from applying to the user on that app, and the browser does not tell the user. See Let people request access. Popup wording comes from shared message templates in User Interaction Hub, so one change can affect several controls.

What the sensitive data window shows​

"Allow" actions show this window when a prompt contains sensitive data. It lists Data types detected, highlights the values and shows an editable copy of the prompt. The user selects Continue with Original, or edits the prompt and selects Continue with Updated. Closing the window cancels the prompt.

  • Allow original prompt with no justification: nothing more is needed.
  • Allow original prompt with optional justification or Allow original prompt with mandatory justification: the user can, or must, give a reason, then selects Submit.
  • Allow prompt after removal of sensitive data: the user removes the data and selects Resubmit and Continue, then Continue once a new scan is clean.

Limit how often a popup appears​

Some Prompt and Justify actions have a Prompt frequency section. The extension uses it only on visit and sign-in controls. The console also offers it on other controls, but there the popup shows every time. See How prompt frequency affects the result.

FrequencyThe popup shows
First occurrence (default)Only the first time.
Every occurrenceAgain each time the Self-resolution period (24 hours, 7 days or 30 days) has passed since it last showed.
Number of occurrencesOccurrence count times (1 to 10), with the Self-resolution period between showings.

Start a Quilly follow-up​

Activate Agent starts a Quilly agent (Quilly is QuilrAI's AI coach for end users) that messages the user in Slack or Microsoft Teams. It needs an enabled Slack or Microsoft Teams integration and at least one organizational policy.

  1. Under Action, choose Policy Explainer Agent, AI Coach Agent, Password Hygiene Agent or Vulnerability Explainer Agent.
  2. Choose the Policy the agent answers from, and optionally a Tone.
  3. Select Generate agent summary.

Conversations appear on the Quilly tab of Users. If Activate Agent is not offered, it is not available for your organization.

Choose Monitor or Enforce​

New controls start in Monitor.

  • Monitor: the user sees nothing, a finding is recorded in Findings & Interactions, and only Remediation Action runs.
  • Enforce: Just in Time, Remediation Action and Activate Agent actions all run.

The extension chooses the control to act on before it checks the mode, so a Monitor control can stop an Enforce control from acting. See A Monitor control can silence an Enforce control.

Set how serious a match is​

Criticality is Very Low, Low, Moderate, High or Very High. New controls start at High. Findings show it, but it does not set the finding's severity. When several controls match, the extension acts on one: the highest risk behavior rank wins, then the highest criticality. See Which control wins when several match.

How System controls differ from Custom controls​

System controls come from QuilrAI and carry the System tag. All start in Monitor, and some start switched on (see System controls compete too). Custom controls are the ones your admins create.

Both kinds can be edited, switched on or off, and duplicated. A copy is always Custom. Only Custom controls can be deleted.

Changing a System control stops QuilrAI updates to it

Once you edit a System control, change its mode, or switch it on or off, QuilrAI stops updating it for your organization.

Use case reference​

The main use cases, grouped by risk behavior, with the mandatory conditions each one adds. "(editable)" marks a value you can change.

  • Blocked AI Applications and Blocked SaaS Applications
    • AI application blocked by organisation and SaaS application blocked by organisation: User Action is Visit; Block status is Yes (editable).
  • Unapproved AI Applications Used: both add Category is Generative AI and Status is Unapproved.
    • A user is visiting an unapproved AI app: User Action is Visit.
    • A user is using an unapproved AI app: Authentication Type is Sign Up or Login (editable).
  • Visiting SaaS app
    • A user is visiting an application: User Action is Visit. Needs an additional condition.
  • SignIn and SignUp to Applications
    • A user signing in and signing up to an application: Category is Generative AI; Authentication Type is Sign Up or Login.
  • Credential Sharing
    • A user is using shared credentials: Authentication Type is Login (editable).
  • Password Reuse: both add Authentication Type is Login (editable), and both need an additional condition.
    • When user is reusing IDP password: Source Application IdP status is Yes.
    • When user is reusing a critical app password: Source Application Criticality is Critical.
  • Compromised Password
    • A user is using a compromised password: Compromised password is Yes; Authentication Type is Login (editable).
  • Weak Password
    • A user is using a weak password: Password strength is less than equal to Weak threshold (editable); Authentication Type is Login (editable).
  • Users Leaking Sensitive Data: both add Sensitive is Yes.
    • A user shares sensitive data with an AI application: Category is Generative AI.
    • A user shares sensitive data with non-AI application: Category is not Generative AI.
  • Users Posting Data: both add Sensitive is Yes or No.
    • A user posting data with an AI application: Category is Generative AI.
    • A user posting data with a non-AI application: Category is not Generative AI.
  • Users Posting Data to a Chatbot: both add User Action is Post and Chatbot is Yes.
    • A user posting data into a chatbot
    • A user posting sensitive data into a chatbot: Sensitive is Yes.
  • Users Copying Data: all add User Action is Copy.
    • A user is copying to clipboard: needs an additional condition.
    • A user is copying data from a business-critical application.: Criticality is Critical.
    • A user is copying sensitive data: Sensitive is Yes; Category is Generative AI.
  • Users Pasting Data: all add User Action is Paste.
    • A user is pasting from clipboard: needs an additional condition.
    • A user is pasting data into an unapproved application: Status is Unapproved.
    • A user is copy-pasting data from one app to another: Source Application Name and Application Name (you choose both).
    • A user is pasting sensitive data: Sensitive is Yes; Category is Generative AI.
  • Users Uploading Data: all add User Action is Upload.
    • A user is uploading data: needs an additional condition.
    • A user is uploading sensitive data: Sensitive is Yes; Category is Generative AI.
    • A user is uploading a labeled document: Label Name (you choose the label).
    • A user is uploading an encrypted file: Encrypted is Yes (editable).