Oddest ChatGPT Leaks Yet: Cringey Chat Logs Found in Google Analytics Tool

Why It Matters

The leak exposes sensitive user inputs to third‑party sites and search engines, eroding trust in ChatGPT and raising potential regulatory scrutiny over data privacy practices. It underscores the need for stricter safeguards when AI models interact with external search services.

Summary

Researchers Jason Packer and Slobodan Manić discovered that hundreds of ChatGPT user prompts – some over 300 characters – were surfacing in Google Search Console (GSC) after a bug caused the AI to prepend a ChatGPT URL to every query that triggered a Google search. Their testing suggests OpenAI may have been scraping Google search results and inadvertently routing raw user prompts to Google, exposing private conversations to any site ranking for the tokenized URL "openai index chatgpt". OpenAI confirmed awareness of the issue and said it had resolved a glitch that temporarily misrouted a small number of queries, but declined to verify whether it had been scraping Google or to quantify the leak. The incident follows earlier ChatGPT privacy scandals, highlighting ongoing challenges in safeguarding user data in AI‑driven web searches.