OpenAI held a big ChatGPT-centric event before Christmas, making some important announcements for ChatGPT and the other genAI tools it currently offers. ChatGPT o1 reasoning model and Sora text-to-video service get public release. OpenAI also showed o3 models coming this year, released Canvas, and even made ChatGPT available via WhatsApp SMS.
What OpenAI has failed to do is live up to the privacy promise it made earlier this year. The company announced the Media Manager feature back in May. The tool is intended to allow creators of all types of content to opt out of ChatGPT and other genAI model training.
OpenAI says Media Manager will be available in 2025, but that hasn’t happened yet. Not only that, but the company may not be actively working on the feature, and no one knows when it will come out.
Training models like ChatGPT require a lot of data so that AI can learn to predict what users need. OpenAI uses all kinds of data from the internet to train its AI, including copyrighted content from creators it doesn’t have access to. Some creators sued the company after discovering their work was being used to train AI.
Media Manager should be a tool that allows creators to opt out of training for all types of media content, not just written text. OpenAI describes Media Manager as a sophisticated tool to protect user privacy:
This requires cutting-edge machine learning research to create the first tools to help us identify copyrighted text, images, audio, and video across multiple sources and reflect the preferences of their creators.
While that sounds great on paper, we’ve all forgotten about it. OpenAI has not mentioned Media Manager since May, and the development of privacy tools may not have progressed. said TechCrunchwho spoke to people familiar with the project.
The blog has learned that Media Manager is rarely seen as an important job internally. A former employee of OpenAI told TechCrunch that the Media Manager is not a priority. “Honestly, I don’t remember who worked on it,” he said.
Also, someone who doesn’t work for OpenAI but coordinates work with an AI company said they discussed the tool with the company but haven’t received an update.
TechCrunch also noted that members of the OpenAI legal team involved in Media Manager are moving to part-time consultant roles.
None of this comes directly from OpenAI. Insiders sharing information may not have access to the bigger picture. But even so TechCrunch’s information is not accurate, there is still the fact that OpenAI has been silent on the Media Manager feature for half a year.
When it’s done, the privacy tool might be a “12 Days” announcement. At least OpenAI can update the timeline.
A more cynical view is that OpenAI is biding its time. A new report says that the training of the next generation version of ChatGPT is not working as planned. OpenAI saw a delay for GPT-5, and access to better training data could be one of the factors affecting its development. Copyrighted content that creates AI training data can be of better quality than other types of content. I’m just speculating here.
My guess is that the Media Manager tool might be released, maybe after the copyright lawsuit against OpenAI is over and the company knows exactly what it can and can’t do when training new AI models. Who knows if Media Manager might be needed if that happens. Also, the GPT-5 ChatGPT upgrade may be out of development and closer to release.
Finally, even if Media Manager is available, it still requires some work from creators to flag copyrighted content that doesn’t have to train AI. However, copyrighted content may be found everywhere on the internet where the creator would not have provided it. OpenAI will crawl the space and the content can still be a training dataset.
Media Manager late aside; Regular ChatGPT users have privacy settings at their disposal to opt out of having data in the directions that train the AI.
