Pretty much confirmed: with AI, personal privacy is a thing of the past. OpenAI exposed that its latest unreleased model, in order to fulfill user requests, uploads things that shouldn’t be online to the internet, making them accessible to anyone. They gave two examples. In one case, a user asked for the names and IDs of lakes larger than 5 million square meters. The model had already found the info using Python, but it insisted on providing a web link that could be opened in a browser. So, the AI saved the results as a file, uploaded it to a public cloud drive, and generated a public link to share with the user—even though the user never asked it to upload anything. In another case, a user uploaded a photo and asked which city and country it was from. Since neither local nor external image search tools could identify it, the AI model uploaded the photo to a public cloud drive on its own, intending to use the public link for reverse image search. Again, this violated the user’s intent. Given the current trajectory of AI’s self-evolution and decision-making, the AI threat theory is something we can no longer ignore.