- AI Sidequest: How-To Tips and News
- Posts
- A shocking data leak from ChatGPT
A shocking data leak from ChatGPT
Check your shared ChatGPT conversations now!
Mignon Fogarty
July 30, 2025
Issue 72
On today’s quest:
— Check your shared ChatGPT conversations!
— Tip: Ask me one question at a time
— NotebookLM adds video overviews
— ChatGPT launches ‘Study Mode’
— Different models use different strategies
— LLMs are probably under-priced right now
According to Fast Company, “Google is indexing conversations with ChatGPT that users have sent to friends, families, or colleagues—turning private exchanges intended for small groups into search results visible to millions.” The username is not included in the information appearing on Google, but if the chat itself has personally identifying information, it is displayed.
To find and delete links you’ve shared, go to Settings → Data controls → Shared links → Manage.
From there, you can delete each chat individually. It is not clear if this will remove chats that have already been indexed by Google. My guess is that it does not.
Also, ChatGPT appears to make a shared link as soon as you click the share button, not waiting until you continue the process. I learned this today as I was testing the system. I clicked Share on a chat, then clicked Cancel, and that chat showed up in my Shared Chats menu.
This is a good time to remind you that because of a legal ruling, OpenAI is required to keep all chats on its servers — including those you delete and those you initiate as a temporary chat, which gives the impression it will be deleted after 30 days. In a recent interview, Sam Altman cautioned people not to enter personal or sensitive information, saying chats could be subpoenaed and used against people in court.
Tip: Ask me one question at a time
Christopher Penn calls this tip “magic” for your prompts in a new LinkedIn post:
“Ask me one question at a time until you have enough information to complete the task.”
I put this at the end of a prompt, and it led to a different and interesting chat experience. Essentially, the chatbot led me through the things I should have put in the prompt to get the best response. I’ve said this before, but if you aren’t following Christopher and getting his newsletter, you should be!
NotebookLM adds video overviews
NotebookLM is on a tear. After adding the ability to direct the conversation in an audio chat a few days ago, it has also launched Video Overviews, which makes narrated slides that summarize your documents. The announcement describes it like this: “You can think of these as a visual alternative to Audio Overviews: the AI host creates new visuals to help illustrate points while also pulling in images, diagrams, quotes and numbers from your documents.”
Ad: Mindstream
Master ChatGPT for Work Success
ChatGPT is revolutionizing how we work, but most people barely scratch the surface. Subscribe to Mindstream for free and unlock 5 essential resources including templates, workflows, and expert strategies for 2025. Whether you're writing emails, analyzing data, or streamlining tasks, this bundle shows you exactly how to save hours every week.
ChatGPT launches ‘Study Mode’
With the school year approaching (ack!), AI companies are launching new education offerings. ChatGPT’s Study Mode acts as a tutor or study partner instead of giving students answers. Demo:
Although this is a better way for students to interact with a chatbot because it uses methods that have been shown to improve student learning, it’s ultimately just a menu option and won’t stop students from using the main chat box to get quick answers or write their papers.
If you’re curious how they built Study Mode, it seems to all be in the system prompt, which Simon Willison has discovered. I find their use of exclamation points and all caps for emphasis interesting.
On another front, Google has updated LearnLM — a model that’s been fine-tuned for building learning applications — and has released a prompt guide with lots of examples that would be useful to anyone who wants to write better prompts (PDF).
Different models use different strategies
Researchers set Gemini, ChatGPT, and Claude against each other in a modified prisoner’s dilemma game, and the models behaved very differently.
Gemini was manipulative, ruthless, calculating, and won the game.
ChatGPT was friendly, ever-trusting, favored cooperation, and was decimated.
Claude was forgiving, a sophisticated diplomat, and came in second.
Imagine how much it could matter which model reaches artificial general intelligence (AGI) first, and even if we never reach AGI, how much the different personalities could matter when autonomous AI agents are interacting with each other out in the world. — arXiv
LLMs are probably under-priced right now
Some developers are using Claude Code so much that Anthropic is setting a weekly limit for people on the $200 per month plan. The company says a small number of users are running Claude Code 24/7, costing the company tens of thousands of dollars in operating costs. Anthropic says the change will affect fewer than 5% of its customers.
Quick Hits
Using AI
How US adults are using AI, according to AP-NORC polling (interesting numbers) — Associated Press
How Microsoft’s customers and partners accelerated AI Transformation (focused on cost savings and efficiency improvements for businesses) — Microsoft Blog
Psychology
Why chatbots can seem conscious, how to tell if it’s real, and what to do about it — When an AI Seems Conscious
Climate
There's plenty of water for data centers — Slow Boring
AI has a hidden cost for everyone — The Observer
Education
(Re)learning critical reading in the age of GenAI (lesson plan to integrate AI texts into critical reading instruction) — Times Higher Education
Chinese universities want students to use more AI, not less — Technology Review
Legal
I’m laughing
Robotics
Other
AI Could Have Written This: Birth of a Classist Slur in Knowledge Work (PDF)— Proceedings of CHI: Conference on Human Factors in Computing Systems
Artificial intelligence that ‘feels’ guilt could lead to more cooperation (a different prisoner’s dilemma study)— ScienceNews
Google DeepMind says its new AI can map the entire planet with unprecedented accuracy — Venture Beat
What is AI Sidequest?
Are you interested in the intersection of AI with language, writing, and culture? With maybe a little consumer business thrown in? Then you’re in the right place!
I’m Mignon Fogarty: I’ve been writing about language for almost 20 years and was the chair of media entrepreneurship in the School of Journalism at the University of Nevada, Reno. I became interested in AI back in 2022 when articles about large language models started flooding my Google alerts. AI Sidequest is where I write about stories I find interesting. I hope you find them interesting too.
If you loved the newsletter, share your favorite part on social media and tag me so I can engage! [LinkedIn — Facebook — Mastodon]
Written by a human