Skip to main content
The input field is used to enter your request or work instruction for the connected AI. Above the input field and the heading “How can I help?”, you can always see which assistant is currently active.
Home > Input Field
1

Enter the desired text.

2

Click the arrow on the right side of the input field. Alternatively, press the “Enter” key on your keyboard.

3

If you want to supplement your input with a document, click the paperclip icon on the right side of the input field. A window will open to select the desired document. When you send the message, VARIOS AI stores attached documents in the uploads/ folder of the files in the chat.

4

After confirming your request, the circle with the progress indicator shows that a response is being generated for you.

Security

To ensure no confidential data is unintentionally transmitted, there is the so-called DLP module (Data Loss Prevention). It detects data such as names, email addresses or phone numbers in your message and replaces it before the message is sent to the AI. If DLP is active for the selected assistant, the DLP icon with a check mark appears on the right in the input field. Whether and how DLP applies is defined by the settings of the assistant and the model: If DLP is enforced for the assistant or the model, it counts as enforced for the chat.

Review findings before sending

With the Optional setting, VARIOS AI checks your message when you send it. If confidential data is detected, the text becomes read-only and the findings are highlighted. Below it, the hint “The DLP module has detected potentially confidential data. It is anonymized automatically before sending. Click a finding to change its protection status.” appears.
1

Review the findings

Findings highlighted in green with a shield icon are protected and are replaced before sending. Hover over a finding to see its type, for example PERSON or EMAIL_ADDRESS.
2

Unprotect individual findings

Click a finding to exclude it from anonymization. The finding is highlighted in yellow and shown with a crossed-out shield. If the same value occurs several times, all occurrences change together. Clicking it again protects the value again (Anonymize again).
3

Unprotect all findings (optional)

The Unprotect all findings switch excludes all findings of the message at once. Turn it off again to protect all findings again.
4

Send

Click Send. Protected findings are replaced, unprotected ones are sent to the AI unchanged. If you have excluded at least one finding, VARIOS AI asks under Confirmation required whether you really want to send the unprotected values. Confirm with Yes, send.
Cancel ends the review. Your text is then back in the input field for editing.
An unprotected value stays unprotected for the entire chat: the AI receives the real value in this message and in all following messages of the chat, and the value is not highlighted again in this chat. This cannot be undone after sending. The decision is recorded in the compliance log.
If a message contains data that must not be sent according to your organization’s rules, the hint “This message contains data that must not be sent. Please adjust the text.” appears. Adjust the text and send it again.

Protected data in the chat history

  • Your messages show the real values. Protected values are highlighted in green. Hover over one to see under Replaced with the value the AI received.
  • AI responses also show the real values. Replaced values in the response are translated back for this.
  • The Show replacements link below a message shows the entire chat the way the AI received it. Show original switches back.
  • The shield icon on a response indicates that DLP changes can influence the response of the AI.
If you edit a sent message using the pencil icon, it goes through the same review when you send it again.