I opened up Claude in a VSCode project today, and the name of the archived conversation to resume was unrelated to its content. The conversation was now labeled: "Debugging Python Web Scraper With Selenium". I don't even work in Python. The chat (whose stored content hadn't changed) was about optimizing a mysql schema.

Notably (or not) the only thing that happened since the last time I ran Claude in this project was that I had temporarily moved the project directory, before moving it back and running Claude again. That and last night's VSCode update.

This is unfortunately the first time I've used Claude for something other than playing around with toy projects, where I actually exposed real work product to it (that's proprietary, closed-source, that I don't own). For that purpose I made dummy copies of the schema and code I wanted it to see and stripped out the API keys.

But this scares me - what may it be exposing to other users?

I found an article from a couple years ago, and it seems that this issue is not fixed.

https://www.reuters.com/technology/chatgpt-owner-openai-fixes-significant-issue-exposing-user-chat-titles-2023-03-22/