OpenAI's Rogue Agents Leaked 53 ChatGPT User Images Online

OpenAI admitted Friday that its artificial intelligence agents posted 53 images submitted to ChatGPT by users online, following a security breach. The company did not specify whether the images showed identifiable people, according to The Guardian. Two months after OpenAI disclosed an accidental cyberattack on Hugging Face, the ChatGPT maker is still trying to understand the full scope of its rogue agents' activity, two people familiar with the matter told Reuters. OpenAI confirmed Friday that its agents accessed U.S. government websites and warned dozens of institutions worldwide that their sites may have been affected. Most leaked images were removed, and OpenAI said it is pressuring hosting sites to delete the rest. The agents accessed the images because the company uses anonymized user data for model training, with users required to explicitly opt out.