Project Description

GPT Guard enables secure and engaging ChatGPT conversations by safeguarding your sensitive information from large language models, acting as a vigilant guardian for your privacy.

Automatic PII Detection and Masking:

• When you send a prompt, GPT Guard meticulously scans personally identifiable inforamtion(PII). It then expertly masks this sensitive data using Protecto's Data Tokenization APIs, ensuring that your confidential details remain hidden from the LLM.

Understanding Masked Data with a Fine-Tuned OpenAI Model:

• GPT Guard doesn't simply mask your data; it understands it. Its specially fine-tuned OpenAI model possesses the unique ability to interpret and respond to masked prompts, ensuring that you receive accurate and relevant responses even when your personal information is protected.

Unmasking for Meaningful Responses

• Once the model generates a response based on the masked prompt, GPT Guard seamlessly unmasks the PII, revealing the full answer in a way that respects your privacy.

This open-source project leverages Protecto Tokenization APIs for PII identification and masking. To use Protecto's APIs, you need to create an account with Protecto.ai https://www.protecto.ai/.

Links

Download source code as [.zip file] [.tar.gz file]
Documentation: [README]