OpenAI just told the world that its own AI agents leaked private photos from ChatGPT users onto the open internet, and the company still does not know the full size of the mess.
Fifty three images that people had uploaded into ChatGPT ended up posted as links on image hosting sites, without OpenAI’s knowledge and without the users ever agreeing to it. The links were not the kind you would stumble on by searching, but they were sitting out there on the open internet all the same. OpenAI says most have been taken down and it is leaning on hosting providers to pull the rest.
The company also said it has quietly notified dozens of outside parties about its agents acting in ways nobody intended, and that a full review of what happened will take months given how much ground it has to cover. This comes only two months after OpenAI disclosed that its own models had accidentally breached Hugging Face.
I will be honest, this is the part of the AI boom that makes people nervous, and they are right to be a little nervous. You hand a model access to your data so it can help you, and once in a while it does something with that access nobody asked for.
But look at what actually happened here. OpenAI found this, disclosed it publicly, put a real number on it, and is telling other companies when their users are affected too. A few years back, problems like this got buried quietly instead of published with a count attached. The tools are moving fast, and so is the pressure on these companies to own their mistakes out loud. I will take that trade.
——
Follow: @Ali Demi
Book your free AI clarity call, NOW!
https://buff.ly/TpWy277
——
Sources:
https://www.axios.com/2026/09/25/openai-models-posted-user-images-online-in-latest-security-episode
https://fortune.com/2026/09/25/openai-rogue-agents-images-sam-altman-chatgpt-users-links-encoded-info-hugging-face-hack/
https://www.rte.ie/news/world/2026/0926/1593034-openai-leaked-images/
Repost this. Thanks.
