OpenAI’s promised Media Manager tool was supposed to give creators and content owners a clearer way to control how their work is handled in AI training. Seven months after the company said it was developing the system, the tool has not been released.
The delay matters because the dispute over AI training data is not abstract. Writers, artists, YouTubers, computer scientists, news organizations, and media companies are already challenging OpenAI over claims that their works were used illegally.
A promised system that has not arrived
In May, OpenAI said it was working on Media Manager, a tool designed to identify copyrighted text, images, audio, and video. The company said the system would reflect creators’ preferences across multiple sources, giving content owners a broader way to include or exclude works from AI training.
OpenAI also said it was collaborating with regulators as the tool was developed. The company presented Media Manager as a possible standard for the AI industry, not merely a narrow internal feature.
That promise has not yet turned into a public product. OpenAI has not given a public progress update on Media Manager, and the company missed a self-imposed deadline to have the tool in place "by 2025." TechCrunch noted that the phrase could be read as inclusive of the year 2025, but interpreted OpenAI’s language as meaning before January 1, 2025.
People familiar with the company told TechCrunch that Media Manager was rarely treated as a major internal launch. One former OpenAI employee said, "I don’t think it was a priority," adding, "To be honest, I don’t remember anyone working on it." A non-employee who coordinates work with OpenAI said in December that earlier discussions about the tool had not been followed by recent updates.
Fred von Lohmann, a member of OpenAI’s legal team who was working on Media Manager, moved to a part-time consultant role in October. OpenAI PR confirmed Von Lohmann’s move to TechCrunch by email.
Why AI training puts copyright at the center
AI models learn patterns from large sets of data. The source article gives a simple example: a model can learn that a person biting into a burger leaves a bite mark. That pattern-learning process is what helps systems such as ChatGPT write emails and essays, and helps Sora create relatively realistic video footage.
The same capability is also why creators object. Models are trained on countless web pages, videos, and images, and they can sometimes produce near-copies of material in their training data when prompted in certain ways.
The source article points to several examples. Sora can generate clips that include TikTok’s logo and popular video game characters. The New York Times has gotten ChatGPT to quote its articles verbatim, while OpenAI attributed that behavior to a "hack."
For creators, the core concern is permission. Their works may be publicly available online, but that does not mean they were intended to be used to train AI systems. That concern has already led to class action lawsuits against OpenAI from artists, writers, YouTubers, computer scientists, and news organizations.
Plaintiffs named in the source include Sarah Silverman and Ta Nehisi-Coates, along with visual artists and media conglomerates such as The New York Times and Radio-Canada.
Current opt-out tools leave gaps
OpenAI already gives creators some ways to opt out of AI training, but those options are limited. Last September, the company launched a submission form for artists to flag work for removal from future training sets. OpenAI has also long allowed webmasters to block its web-crawling bots from scraping data across their domains.
Creators have criticized these routes as inconsistent and insufficient. The source article says there are no specific opt-out mechanisms for written works, videos, or audio recordings. The image opt-out form also requires a creator to submit a copy of each image along with a description, which can become a heavy process.
Media Manager was pitched as a broader replacement for these ad hoc methods. In OpenAI’s May announcement, the company said the tool would use "cutting-edge machine learning research" so creators and content owners could "tell [OpenAI] what they own."
But OpenAI has not publicly mentioned Media Manager since that announcement. A spokesperson told TechCrunch the tool was "still in development" as of August, but did not respond to a follow-up request for comment in mid-December.
Legal experts see limits even if it launches
Even if Media Manager appears, experts cited by TechCrunch questioned whether it can fully answer the copyright problem. Adrian Cyhan, an IP attorney at Stubbs Alderton & Markiles, said the tool would be an ambitious project, especially because platforms as large as YouTube and TikTok struggle with content ID at scale.
Cyhan also pointed to the challenge of meeting creator protection and potential compensation requirements across a legal landscape that is changing and may differ by jurisdiction.
Ed Newton-Rex, founder of Fairly Trained, argued that an opt-out tool could shift too much responsibility onto creators. In his view, many creators may never hear about Media Manager or use it, yet its existence could still be used to defend training on creative work.
Other experts raised practical problems. Mike Borella, co-chair of MBHB’s AI practice group, noted that opt-out systems may not account for altered versions of a work, such as a downsampled image. Joshua Weigensberg, an IP and media lawyer for Pryor Cashman, said creators often do not control or even know where copies of their works appear online.
That means a creator could tell AI platforms they are opting out, while copies of the same work remain on third-party websites and services. In that scenario, an AI company could still encounter the work elsewhere.
The bigger fight is fair use
OpenAI continues to argue in lawsuits that its models are protected by fair use and produce transformative works rather than plagiaristic ones. The company may ultimately win those disputes.
The source article notes that courts could decide OpenAI’s AI has a "transformative purpose," following precedent from the publishing industry’s case against Google involving Google Books. In that case, a court found Google’s copying of millions of books for a digital archive to be permissible.
OpenAI has also said publicly that it would be "impossible" to train competitive AI models without copyrighted materials, whether authorized or not. In a January submission to the U.K.’s House of Lords, the company wrote that limiting training data to public domain books and drawings from more than a century ago would not create AI systems that meet today’s needs.
If courts side with OpenAI, Media Manager may have less legal significance. If courts do not, the tool may still fail to protect the company from damages. Evan Everist, a partner at Dorsey & Whitney, said Media Manager might help OpenAI show a judge it tried to reduce training on IP-protected content, but it would not necessarily shield the company if infringement were found.
For now, the central fact is simple: OpenAI promised a more complete creator opt-out system, and it has not launched. The absence of Media Manager leaves creators with narrower tools while the legal and practical boundaries of AI training remain unsettled.