Check immediately with rules
Regular expressions and keywords find sensitive data with predictable formats first. Clear matches can be decided without waiting for AI analysis.
Data Loss Prevention
Generative AI moves information through conversations, documents, code, and prompts—often with sensitive data embedded in ordinary language. QueryPie DLP quickly screens Korean, English, and Japanese inputs for risk, then identifies what kind of sensitive information appears and its exact text in the inputs that need deeper analysis.
From rule checks to risk screening and sensitive-data extraction
Checks input with regular expressions, screens risk with an ELECTRA-based model, and extracts sensitive data with an SLM-based model.
Screen risk with an ELECTRA-based model
A lightweight classifier reads text that did not match the rules and calculates a risk score. High-risk and ambiguous inputs are selected for closer review.
Extract sensitive data with an SLM-based model
For input that needs precision analysis, the model finds what kind of sensitive information appears and its exact text. A separate program checks the source to calculate its position.
Fast risk screening
A lightweight ELECTRA-based model quickly evaluates sensitive-data risk in text that rules alone cannot assess. It identifies inputs that need closer review, even at high volume.
Quickly filters risk signals from high-volume inputs.
Fast risk screening
ELECTRA-based DLP
Examines text that patterns and formats alone cannot classify.
Scores the likelihood that the input contains sensitive data.
Identifies high-risk or ambiguous inputs for closer review.
ELECTRA-based risk screening demo
Fast risk-screening capabilities
Test sentence
The received file contains the following:
Connection IP 198.51.100.227.
Risk screening result
0.999
Block decision
Enter a sentence to see the lightweight model's risk score and block decision.
Precise sensitive-data extraction
An SLM-based model analyzes context to identify the types of sensitive data and their exact text in the input. This makes it clear which parts need protection.
Input document
Contact Kim Minsu's
minsu.kim@example.com
Phone 010-1234-5678
Detection results
- ContactKim MinsuPERSON NAME
- minsu.kim@example.comEMAIL ADDRESS
- Phone010-1234-5678PHONE NUMBER
Enter a sentence to see the kinds of sensitive data found and the exact text for names, email addresses, phone numbers, and more.
SLM-based DLP
Distinguishes names, account details, credentials, and other sensitive data.
Shows how the sensitive information is written in the input.
Reads the surrounding text to assess whether an expression outside a fixed pattern is sensitive.
Use the DLP models on Hugging Face
Explore QueryPie's multilingual sensitive-data detection models, model cards, and usage resources on Hugging Face.
Stop Thinking.
Start Transforming.
