GDPR Compliant ChatGPT Playground

ChatGPT is an AI based conversational tool developed by OpenAPI on the design principles of Large Language Model (LLM) and a publicly accessible tool. ChatGPT has increasingly becoming popular tool for enabling applications involving interactive and contextual search across corporate companies, profit/non-profit organizations, educational/medical institutions, and researcher’s community to name a few.The corporate companies are abided by GDPR (General Data Protection Regulation) compliance checks and restricted to share confidential, personal or sensitive information’s (hereinafter referred as critical data) into public domains. As ChatGPT server is hosted outside corporate boundaries, to fulfil GDPR compliance check, the corporate companies must have necessary systems in place of any data leaving outside of corporate boundaries.We are proposing a novel solution in identifying GDPR noncompliant DPP (Data Privacy and Protection) entities from the prompt query given to ChatGPT. To achieve this, from the corporate documents we first manually tag critical entities from “named entity tagging tool” in building the corporate specific DPP entities knowledgebase, build custom NER (Named Entity Recognition) model on top of prebuilt corporate specific DPP entities knowledgebase, an ChatGPT playground interface to accept any user’s prompt query, before firing the query to ChatGPT we validate the user’s prompt query against custom NER model to detect if any corporate specific DPP entities are present, warn the user by highlighting the corporate specific DPP entities if present to facilitate user in negating the same, we enabled feedback loop from the user for the highlighted corporate specific DPP entities to improvise the custom NER model and logging all the input prompt queries fired to ChatGPT to enable corporate auditing process by using techniques from Natural Language Processing (NLP), Information Extraction, Information Retrieval (IR) and Custom NER model.