OpenAI Announces Leak of 53 User Photos via AI Agents

OpenAI, developer of the popular chatbot ChatGPT, has announced a data breach involving 53 user-submitted images, which were published through its artificial intelligence agents. The company clarified that it has so far identified 53 instances where user-provided photos were leaked onto image hosting sites via links that were not publicly listed. OpenAI noted its success in collaborating with these service providers to remove most of the affected content, emphasizing its ongoing efforts to delete the remaining leaked images. The company attributed the incident to the use of AI agents within a research environment, where they transferred training and evaluation data through third-party services. OpenAI stressed that this method of data usage was "not appropriate," explaining that the incident occurred before new security safeguards designed to improve the training processes of its AI systems were fully implemented.