Chat Interface Screening for Sensitive AI Prompt Submission

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Generative AI technologies accessed from endpoint devices beyond organizational firewalls risk exposing sensitive information, compromising confidentiality and proprietary data.

Innovation Solution

Implementing a computer-implemented method with an endpoint agent that scans and determines if text intended for transmission to a generative AI tool contains sensitive information, automatically canceling transmission if sensitive, and providing notifications or alternatives.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If text transmission to generative AI tool is enabled, then access to AI functionalities is improved, but sensitive information disclosure risk increases

Engineering Contradiction:

Improveaccess to generative AI functionalitiesVSAvoidsensitive information disclosure risk

Core Design Contradiction:

ProductivityVSObject-affected harmful factors

Solution Approach 1:

The scanning engine performs preliminary scanning of text content before transmission to the generative AI tool. This preliminary action identifies sensitive information in advance, allowing the system to block transmission of harmful content while preserving access to safe content, thus resolving the contradiction between productivity and security risk

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The scanning engine acts as an intermediary component between the user interface and the generative AI tool. It mediates the text transmission process by filtering and inspecting content, enabling the system to allow beneficial AI access while preventing sensitive information disclosure through this intermediate security layer

Inventive Principle:

Principle #24Intermediary (Mediator)

2Object-affected harmful factors

If text transmission is blocked, then sensitive information disclosure is prevented, but access to generative AI is restricted

Engineering Contradiction:

Improvesensitive information disclosure preventionVSAvoidaccess to generative AI functionalities

Core Design Contradiction:

Object-affected harmful factorsVSProductivity

Solution Approach 1:

The scanning engine applies local quality analysis by examining specific characteristics of text content to distinguish between sensitive and non-sensitive information. This localized inspection allows the system to block only harmful transmissions while permitting beneficial AI interactions, resolving the contradiction through selective filtering rather than blanket blocking

Inventive Principle:

Principle #3Local quality

3Object-affected harmful factors

If scanning for sensitive information is performed, then information security is improved, but transmission delay increases

Engineering Contradiction:

Improveinformation securityVSAvoidtransmission delay

Core Design Contradiction:

Object-affected harmful factorsVSLoss of time

Solution Approach 1:

The scanning engine performs partial scanning by focusing on identifying sensitive information patterns rather than complete content analysis. This partial action approach provides sufficient security protection while minimizing processing time and transmission delay, resolving the contradiction between information security and time loss

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20260003999A1Preventing the disclosure of sensitive information in attempts to Utilize Generative Artificial Intelligence (AI) through a Chat-Style Interface

Publication Date: 2026.01.01 PROOFPOINT INC

AI summary

A computer-implemented method is disclosed that includes receiving a notification that a submit option has been selected on a chat interface to a tool (e.g., a generative artificial intelligence (AI) tool) hosted at a remote network destination, cancelling the transmission of text from the chat interface to the generative AI tool is canceled in response to the notification. Text, initially intended to be submitted to the generative artificial intelligence tool from the chat interface, is obtained and scanned to determine whether the content of the text contains sensitive information. Typically, this determination is made based, at least in part, on the content of the text. Depending on an outcome of the determination, the computer either enables or prevents transmittal of the text from the chat interface to the generative artificial intelligence tool. If enabled, the computer typically and automatically reselects the submit option (without further involvement from the user).