A French-language YouTube monologue warning about a major ChatGPT data leak in which at least 100,000 private conversations (the speaker cites ~110,000) were publicly exposed after being indexed by Google and archived on archive.org. The speaker explains the removed "share chat" feature that generated unprotected links (no password/authentication), how Google crawlers and the Wayback Machine made those links discoverable, and gives concrete examples of sensitive disclosures (intimate/health data, confidential contracts, legal strategies to displace indigenous people, political dissent). The video warns companies and individuals about the irreversible loss of control when private documents are given to AI services and notes that some firms already ban employee use of such AI tools.
Questions this video answers
The speaker cites about 110,000 (also refers generally to "more than 100,000").
Google (search/indexing) and archive.org (Wayback Machine).
A ChatGPT "share chat" feature that generated unprotected links with no password or authentication.
Hello, my friends. More than 100,000 private conversations on GPT Chat have been leaked. And they were even worse than leaked. They were indexed on Google and even worse, they were indexed on the archive.org website. So what is this about? There is a feature on GPT Chat that was removed precisely following this scandal that allowed you to share a chat. That means you are chatting on GPT Chat and what you can do is you have a button that allows you to generate a link and you can send this link to whoever you want. The person clicks, opens it and sees the exchange. This link has no protection. It is a link that was generated. It has no password, no authentication, no anything as long as you share it. It is like a bearer check in the old days, so whoever has this link can access the conversation. A lot of people also did. What they did was not start a new conversation. That means continuing on a thread that was sometimes several days old, clicking Share all when they thought they were maybe just sharing the answer that the system had just given them. What happened? Google, which is very good at creating links. Where does Google find them? It's not that people were necessarily posting them on forums or whatever. That's not even it. It's that, for example, you exchange a link with someone on Gmail, who owns Gmail. It belongs to Google. So it's a tool that will potentially detect links. And then it can send its crawlers, index them, what's the password for them, and add them to the Google index. Then behind it, what do you have? You have the Wayback Machine, as we call it. Wayback Machine. It's a very interesting system. It's archivearchive.org org it's the site. So if you want, it's a site that lets you go back. That is to say, sometimes you read a version of a page, you say, I'm sure there was something else on this page before. You can put it. Put the link on it and go back. You can also find sites that have disappeared. For example, sometimes you have a site. It happens to me. Sometimes I have saved links on my browser. I click on a link that I saved maybe two years ago, I see that the site no longer exists. No domain, not renewed or anything. I go to the Wayback Machine, which lets me go back and see the article in question. So what happened is that it was a massive indexing. So the damage today we know that there is at least 100,000. I read 110,000, but it could be even more than that. So it's a disaster. What do we see in it. We see some strange things in it. Of course, there are private conversations, intimate letters, health issues and all that. But sometimes we also see some terrible things. For example, a lawyer says, I want to drive an indigenous tribe off some land in the Amazon. Hey, it's an Italian lawyer asking this question. I'm looking for strategies. This tribe doesn't know the value of their land. My client wants to build a hydroelectric dam, but he wants the area to be cleared. What are the methods for doing this? So we now know that there is a hydroelectric dam that is probably planned by European, Italian or other interests in the Amazon, and that there are tribes there and we're going to have to deal with them. So we know that there was someone who wanted to know an Egyptian who wanted to know how to criticize the president of his country, how to write a critical article on Sisi's record. So these are people who could end up persecuted and all that. So this is a huge disaster because it's the first major data leak. The first major AI data leak, in fact. And it's not the first. Well, it's the first, but it's not the last, which is what I was trying to say. It means there will be others. What we also find in this, in this license is that, for example, they are in a company, they receive a contract. The terms of the contract are confidential, but they want a summary. So they're going to give a GPT cat to do a summary. Where is the contract now? Well, it's on the system. What's going to happen? Will it be used to train the model? Will it end up on Google? Will it end up somewhere else? Huh? Gumball Mystery, huh? Once we give, we need to know one thing. Once we give documents information to a system that is not under our control, it's over, huh? It means we have lost, we have lost control of this data. Typical example. You are in a laboratory, huh? A lab, uh, you are developing, I don't know, a new formula or something like that. And then, uh, you need a little help. You upload your formula to chat GPT. That sounds familiar. He will comment on it, he will help you and everything. What will happen. Potentially 20 minutes later, there is a competitor on the other side of the world who asks ChatGPT a question on how to develop such and such a formula potentially, huh? It's integrated, huh? This information that you gave, it's integrated into the system's training and it can be spit out in this form or another to a competitor, you see? So that's what's happening. So today there are. There are many. There are many companies that prohibit it. So as much as there are some who are completely crazy about, about moving towards GPT, chat or other forms of AI, there are also plenty. And I know some who prohibit their employees from using AI. Why? Because they realize the danger behind it. Like Selek, for example. So that's it. I just wanted to tell you to be very careful because it can really cause disasters in your life. If confidential information were to be widely distributed to humanity. That's it, my friends. Thank you very much. This video was made without yard and I'll see you in the next one.
