SafeGPT: Preventing Data Leakage and Unethical Outputs in Enterprise LLM Use

TL;DR

SafeGPT is a comprehensive system designed to prevent data leakage and unethical outputs in enterprise LLM applications through input detection, output moderation, and human feedback.

Contribution

It introduces a novel two-sided guardrail system combining detection, moderation, and human-in-the-loop feedback for enhanced LLM safety in enterprise settings.

Findings

Reduces data leakage risk effectively

Decreases biased and unethical outputs

Maintains user satisfaction

Abstract

Large Language Models (LLMs) are transforming enterprise workflows but introduce security and ethics challenges when employees inadvertently share confidential data or generate policy-violating content. This paper proposes SafeGPT, a two-sided guardrail system preventing sensitive data leakage and unethical outputs. SafeGPT integrates input-side detection/redaction, output-side moderation/reframing, and human-in-the-loop feedback. Experiments demonstrate SafeGPT effectively reduces data leakage risk and biased outputs while maintaining satisfaction.