Google was showing the ChatGPT chats you shared with others
Marco van Hurne
Building AI organizations from strategy to execution | author Machine Learning book of knowledge | teacher | researcher
Good morning y’all ! How I got a surprise this morning. I was reading through my feed when I stumbled across something that made my coffee taste bitter. Apparently Google has been indexing your shared ChatGPT conversations with the rest of the world, and, boy oh boy, are people serving up some spicy content.
More rants after the commercial brake:
So get this. .
Some doofus at OpenAI thought it would be brilliant to make ChatGPT chats you shared with others searchable on Google. Yeah, you read that right. Your confessions to Dr. AI are now part of the public record.
You may have come across this function, and never done anything with that - and in that case: Good luck and no problemo.
But if you have, it would have looked like this:
You shared a chat you thought could be interesting for a friend who has an undiagnosed cases of herpes, you copy the link, smash it in WhatsApp and voila. His problem now.
At least, that’s what you thought!
Turns out that OpenAI forgot to close the door for Google Spiders.
A simple addition to Robots.txt (and a few other tricks) could have resolved the matter, but nay, they probably forgot, or they found they were entitled to a privacy oopsie like Meta had recently (showing all the conversations you had with Meta AI on their portal: read publicly).
All someone needed to do is to use a thing called a Google Search Operator, a link to where ChatGPT stores its chats and followed by whatever they want to snoop on, like this: you type this into the search bar: site:chatgpt.com/share “Marco van Hurne” or “Elon Musk” or “Donald Trump”, whatever you fancy, and bada-bing-bada-boom - instant access to tens of thousands of conversations.
The whole mess started with ChatGPT's "Share" button, you know, that innocent little shit, sitting in the corner that nobody really thinks about. You clicked it, generated a URL, and suddenly your private heart-to-heart with ChatGPT becomes as public as a billboard.
OpenAI tried to warn people (they still do by the way - the service ain’t gone).
They put up a message saying "Anyone with the URL will be able to view your shared chat" but apparently forgot to mention that Google would be indexing these conversations like some sort of archaeologist.
The peeps of Fast Company went digging and they found well over 4,500 publicly indexed chats. The content reads like a soap opera written by people having nervous breakdowns. Mental health struggles, relationship drama, work complaints, business strategies - it's all there for anyone with basic Google skills to find.
The really tragic part was that people were using ChatGPT for resume building and accidentally published their full names including phone numbers, email addresses, and work histories online. One poor schmuck’s entire professional life is now searchable entertainment for bored internet users like me.
And companies got caught with their pants down too.
Think of marketing strategies, internal discussions, proprietary information - all sitting pretty in Google's search results playing hide and seek.
OpenAI's “Chief Information Sharing Officer” eventually showed up to clean the mess. He called it a "short-lived experiment to help people discover useful conversations". Hahahaha…. allow me to translate this: we messed up big time and now everyone's mad at us.
They did not pull the feature (at least not yet in my country), but started begging Google instead to remove the indexed content. That is not cybersecurity folks? Security starts with the cause of it, and not with making pleas to Google for help.
But here's the thing about the internet - it never forgets, even when you desperately want it to. Those conversations might stick around in Google's cache for a long-long-time.
Want to protect yourself?
Here's what I learned from security experts who actually know what they're talking about. Of course, you go check first if your conversations are already out there. Search "site:chatgpt.com/share" plus your name or company, and if you find anything, delete those shared links immediately through ChatGPT's settings, and second, you stop treating AI chatbots like your personal therapist.
Man, these things aren't private confessionals. They are more like speaking into a megaphone at a crowded stadium.
And if you really, really want to spill your guts to a bot, just install Ollama, choose a big model like Llama 4 something and do it locally on your machine.
But if you just can’t help yourself, there is something like ChatGPT's temporary chat mode for sensitive stuff. Did you know those conversations disappear in 30 days, which is still longer than most people's attention spans but better than forever.
You know, this whole farked up situation proves something I've suspected for years. People will share the most intimate details of their lives with a computer program but won't tell their neighbors their first names. It's like we've collectively decided that artificial intelligence deserves more trust than actual humans.
Privacy experts have been screaming about this stuff for months. Treating every AI interaction like it might end up on a billboard is actually good advice. Who knew? This mess hits different when you realize it affects GDPR compliance, corporate data policies, and basically every privacy law that exists. Companies that let employees use ChatGPT for business discussions might have accidentally turned their strategic planning into public entertainment.
The real lesson here however, isn't that OpenAI made a mistake.
It is that most people have no clue how their data moves around the internet (and do not care - “I have nothing to hide”). Everyone is walking around with smartphones that know more about them than their mothers do, but somehow a chatbot sharing feature caught them off guard.
So people, moving forward, you-gotta-treat-every-AI-conversation-like-it-might-end-up-in-your-company's-next-board-meeting-presentation. Keep your personal shit vague man, avoid real names, and maybe don't use ChatGPT to work through your relationship problems unless you want strangers googling your drama later, and remember one thing: The internet doesn't have an undo button, and artificial intelligence companies are still figuring out how to handle user data without creating privacy disasters, and until they sort that out, you’ve got to assume that everything you type is going straight to a bulletin board.
Anyways, them peeps at OpenAI's responded quick to the news, and that shows they understand the gravity of screwing up people's privacy. But trust, once broken, takes years to rebuild. And in the AI world, years might as well be decades given how fast everything changes.
Signing off,
Marco
I build AI by day and warn about it by night. I call it job security. Big Tech keeps inflating its promises, and I bring the pins. I call that balance and for me it is also simply therapy.
Think a friend would enjoy this too? Share the newsletter and let them join the conversation. Google and LinkedIn appreciates your likes by making my articles available to more readers.
To keep you doomscrolling 👇
Adviseur ux & usability
1wSo, the plans you shared with me, to become a dictator for life in some middle American country, are all over the Internet now? Great…
AI Engineer • GenAI / RAG / AI Agents • Shipping LLM solutions at Jaguar Land Rover, NatWest, VW, Wickes
1wThere were some VERY interesting conversations that people documented But was it an oversight or deliberate? 🤔