Highlights: what you see and what you do
What a yellow, red or teal highlight means, how to mask or delete a value, and where to see all highlights in one list.
ChatGPT5Drafting a support promptWhen you type or paste text in an AI tool, an email or a chat, BeeSensible underlines sensitive data in the text field before you send your message. Such an underline is called a highlight. This article explains what the three kinds of highlight mean, what you can do with a highlight and where to see all highlights in one list. A highlight does not change your text or hold your message back, and you decide what to do with the value. The example above this article behaves like the product, so you can try the actions there.
Transcript
On screen: Heading: "What AI does for you", with the label "Your work". A card with three examples that appear one after another. "Speech → report": a voice memo with a waveform becomes a report with the headings "Present", "Discussed" and "Agreed". "Document → summary": a 12-page client file becomes a summary of five points. "Code → done": in an editor half a line of code is completed and accepted with Tab.
Maybe you use AI in your work. You have a conversation written up as a report, a long document summarised or a piece of code finished. AI can save you a lot of time in your work.
On screen: Heading: "What you share goes along", with the label "Your text". A browser window on chatgpt.com. The file "Client file.pdf" goes into the text field, which now reads: "Summarise this: client meeting with Sanne Visser about her application, BSN 123456782. She asks for a payment plan for six months." The name and the number briefly light up in yellow, without a highlight.
To help you, an AI tool needs your text. Sometimes it contains more than you want to share, such as a client's name or a national ID number.
On screen: Heading: "Using AI safely", with the label "BeeSensible". In the same window the BeeSensible icon appears at the top right with "On". The name gets a single line and the BSN a double line.
That is why your organisation is getting started with BeeSensible. BeeSensible works inside the AI tools you already use and helps you spot sensitive data in time. That way you can keep using AI, including for work with client data. In this video you'll see how that works.
On screen: Heading: "Highlights as you type", with the label "Highlights". A browser window on chatgpt.com, with the BeeSensible extension icon in the address bar. In the text field someone types: "Write a reminder to Femke Brouwer that her payment to NL43RABO0198765432 is still outstanding." The name gets a single line and the IBAN a double line. Below the window a legend appears: "Single line: sensitive data, such as a name or address." and "Double line: extra sensitive, such as a BSN or IBAN."
As you type in an AI tool, BeeSensible highlights sensitive data in your text. A single line marks sensitive data, such as a name, and a double line marks extra sensitive data, such as an IBAN.
On screen: Heading: "Masking or deleting", with the label "Actions". Above the name a small window opens with the type "Name" and the level "Sensitive", and the options "[NAME1]" and "Delete". "[NAME1]" is chosen; the text field now shows [NAME1] with a dotted line. The legend adds "Dotted line: handled, for example masked."
When you click a highlight, you choose whether to mask the detail or delete it. Masking puts a placeholder in its place, and anything you have handled gets a dotted line.
On screen: Heading: "Masking when you paste", with the label "Pasting". The text field is empty. Above it, first "BeeSensible is checking what you paste" appears, then "2 sensitive details found in what you paste", with "Nothing is in the text field yet. This is how it goes in when you paste masked." and the text "Dear [NAME1], you can reach me on [PHONE1]." Below are the buttons "Paste masked", "Paste original" and "Cancel". After "Paste masked" that text is in the field, with a dotted line under the placeholders.
When masking when you paste is on for you, BeeSensible first checks what you paste. You see what was found, and Paste masked puts the text into the field with placeholders.
On screen: Heading: "Anonymising a document", with the label "Documents". The BeeSensible side panel slides in on the right, with the tabs "Privacy" and "Documents" and the button "Upload a file". The file "Tenancy agreement.pdf" is dragged to the side panel. A window "Review before anonymizing" opens, with the page on the left, where a name, an address and an IBAN are highlighted, and on the right "3 of 3 selected" with those three items ticked; the IBAN is labelled "Critical". After "Anonymize (3)" the side panel shows "Anonymized documents" with the file "Tenancy agreement_anonymized.pdf", "3 items anonymized", the button "Add to chat" and the line "BeeSensible may miss sensitive information. Review the document before sharing it."
If you want to share a document, you can anonymise it first, provided your organisation offers this feature. You drag it to Documents in the side panel, review what was found and get a copy with that data removed.
On screen: Heading: "A tool that is not allowed", with the label "Notice". The address bar shows chat.deepseek.com. A card appears in the middle of the page with the BeeSensible logo and the word "notice", the label "Not allowed", the heading "DeepSeek is not allowed here" and the text "Your organisation does not allow this tool for work. Use an allowed alternative." Below are "Use instead" with the button "Open Microsoft Copilot", and at the bottom the button "Go back".
If you open an AI tool that your organisation has not allowed, you see a notice with the alternative your organisation offers.
On screen: Heading: "Read your text yourself", with the label "Read it through". Back on chatgpt.com the text field reads: "Thank [NAME1] and mrs a. de wit-bakker in a short email for today's meeting." [NAME1] has a dotted line; "mrs a. de wit-bakker" is not highlighted. The cursor selects "and mrs a. de wit-bakker", the Delete key is pressed and that part disappears. Then a card appears with the BeeSensible logo, "BeeSensible can miss something." and "Always read your text yourself before you send it."
BeeSensible can miss sensitive data, so always read your text yourself before you send it.
On screen: End card: the BeeSensible logo with the title "Getting started with BeeSensible".
What the three highlights mean
Transcript
On screen: Heading: "Highlights as you type", with the label "Highlights". A browser window at chatgpt.com, with the BeeSensible extension icon in the address bar. Someone types in the text field: "Email Femke Brouwer that her refund to DE89370400440532013000 is on its way. She can be reached on +31 6 12345678." Then the name and the phone number get a yellow line underneath and the IBAN a red double line.
As you type in an AI tool, BeeSensible underlines sensitive data in the text field, before you send the message.
On screen: Heading: "What the colours mean", with the label "Colours". A legend appears next to the window: the word "Femke" with a yellow line means "Sensitive", and the word "IBAN" with a red double line means "Critical".
A yellow line means sensitive, such as a name or a phone number. A red double line means critical, such as an IBAN or a national ID number.
On screen: Heading: "Masking or deleting", with the label "The window". A small window opens above the name Femke Brouwer, with the type "Name" and the level "Sensitive". Below that are the options "[NAME1]" and "Delete", and the switch "Apply to all occurrences". The pointer moves to "[NAME1]", the switch "Apply to all occurrences" is turned on and "[NAME1]" is chosen. The text field now shows [NAME1] instead of the name.
Click a highlight to choose the mask label or Delete. With Apply to all occurrences, your choice applies to the same value throughout your text.
On screen: Heading: "A handled detail", with the label "Handled". In the text field, [NAME1] has a teal dotted line. A third line appears in the legend: "[NAME1]" with a teal dotted line means "Handled".
A detail you have handled gets a teal dotted line.
On screen: End card: the BeeSensible logo with the title "Highlights as you type".
Every highlight has a colour and a line style, so you can see at a glance how sensitive a value is. The line style also helps if you find the colours hard to tell apart.
- Sensitive: a yellow, single line. This is data that is sensitive in one situation and not in another, such as a name, an email address or a phone number. With a highlight like this, take a moment to check whether the value belongs in your message.
- Critical: a red, double line. This is higher-risk data, such as an IBAN or a national ID number (the BSN in the Netherlands). Always look at these highlights before you send.
- Handled: a teal, dotted line. You have already dealt with that value, for example by masking it.
The pop-up on a highlight and the side panel also show the level as text, so you do not have to rely on the colour alone.
What you do with a highlight
When you hover over a highlight or click it, a small pop-up appears. It shows the data type, the level and two choices:
- At the top is the mask label, numbered by type, such as
[NAME1]or[IBAN1]. If you choose the label, BeeSensible replaces the value with it. The same value keeps the same number throughout your text, and a second name becomes[NAME2]. The AI tool therefore still knows what type of data was there and which mentions refer to the same person, so its answer stays useful. - Below it is Delete, which removes the value from your text entirely. If that leaves a gap in the sentence, rewrite the sentence yourself.
Which of the two to choose depends on what your question needs:
| Situation | Choose |
|---|---|
| The data is not needed | Delete |
| Your question only needs the type of data | Mask |
| Not sure | Mask |
If the same name or address appears several times in your text, turn on Apply to all occurrences at the bottom of the pop-up before you choose. Your choice then applies to every place where that value appears.
If you are happy to share a value, leave the highlight as it is and send your message. If masking on send is switched on, BeeSensible first masks the values that are still highlighted and then sends the message.
All highlights in one list in the side panel
The pop-up on a highlight is convenient when you want to change a single value. To see all highlights in your text together, open the side panel from the tab with the BeeSensible logo on the right-hand edge of the page. The tab shows a count as soon as BeeSensible has found something. For each highlight, the panel shows the value it found, the data type (such as name or email address), the level (Critical, Sensitive or Handled) and the Mask and Delete buttons. You can also operate the panel with the keyboard.
Which apps show highlights
The extension runs in Chrome and Edge, in AI tools, email, chat and other supported sites. If your organisation also uses the BeeSensible desktop app, you see the same highlights in Outlook desktop, the ChatGPT app, the Claude app and Microsoft Copilot. There you hover over an underline and choose the mask label or Delete, and with Apply to all your choice applies to every place where the value appears. The desktop app makes the change directly in the app's own text field.
ChatGPT 
Claude 
Gemini 
Copilot 
Perplexity 
DeepSeek 
Grok 
Mistral AI 
Meta AI 
Gmail 
Outlook 
Slack 
Discord 
WhatsApp 
LinkedIn 
X 
Facebook If you work with an app that is not listed here and would like BeeSensible to help you there as well, email hello@beesensible.eu.
More about highlights
You do not need the sections below to work with highlights. They give background on what you see and on what happens to your text.
Why do I see a highlight in one app and not in another?
Which data types are highlighted as Sensitive or Critical, and which are not highlighted at all, depends on the detection profile. Your admin sets that profile for each app. As a result, a stricter profile can apply in a public AI tool than in an internal chat, and you may see a highlight in one app that you do not see in another.
Highlights for the names of projects, clients and systems
Your organisation can keep a list of names that only mean something within the organisation, such as the names of projects, clients and internal systems, or internal codes. A custom term like this gets the same highlight as any other data type, Sensitive or Critical, depending on what the detection profile sets for its category. If you mask a custom term, it becomes a placeholder that includes its category, such as [ORG-PROJECT1]. The extension searches your text for the terms on the list on your own device, and the list never goes to the detection model.
When BeeSensible masks by itself
A highlight points out a value to you, and you decide whether to mask it, delete it or simply send it. In AI tools, BeeSensible can also do the masking for you when you paste text, attach a file or send your message, but only if you or your organisation have switched that feature on. If Keep context is on, you see the real values in the AI tool's answer, while the AI tool only receives placeholders. Masking when you paste, attach or send explains how that works.
What happens to your text
To recognise sensitive data, the extension sends your text to BeeSensible's own servers at European providers. They check the text in memory and discard it straight afterwards, so BeeSensible never stores what you type. No external AI service, such as OpenAI, Anthropic or Google, is involved.
Last updated:
