OpenAI has failed to deliver on its promise to create a copyright protection tool by 2025

OpenAI failed to release its promised Media Manager tool by 2025, which would allow content creators to control the use of their work in training neural networks. Media Manager, announced last May, was supposed to identify copyrighted text, images, audio and video.

Image source: hdhai.com

The tool was supposed to help OpenAI avoid legal problems associated with intellectual property infringement, and in general could become a standard for the entire artificial intelligence industry. However, as TechCrunch writes, the development of Media Manager was not initially considered a priority by the company. One former OpenAI employee noted: “I don’t think it was a priority. To be honest, I don’t remember anyone working on this.” Another source close to the company confirmed that there had been discussions about the tool, but there had been no new information related to the project since the end of 2024.

It must be said that recently the use of proprietary content for training AI has repeatedly become the cause of controversy. OpenAI models such as ChatGPT and Sora are trained on huge data sets including text, images and videos from the Internet. This allows AI models to create new works, but they often turn out to be too similar to the original. For example, Sora can generate videos with the TikTok logo or video game characters, and ChatGPT was caught using verbatim quotes from The New York Times articles.

This practice causes a wave of indignation on the part of authors whose works were used without their consent. Class action lawsuits have already been filed against OpenAI from artists, writers and major media companies, including The New York Times and Radio-Canada. Authors such as American actress and screenwriter Sarah Silverman and writer Ta-Nehisi Coates have also joined the lawsuits, accusing OpenAI of illegally using their work.

OpenAI has proposed alternative solutions to the problem, and currently offers several ways for content creators to exclude their work from neural network training. Specifically, in September 2024, a form was launched to request the removal of images from future datasets. The company also has nothing against webmasters blocking their sites from data collection by its bots, for example in the “robots.txt” file. However, these methods have been criticized both for their complexity (removing content from a dataset) and for their imperfections.

Media Manager, on the contrary, was presented as a long-awaited all-in-one solution. In May 2024, OpenAI said it was working with regulators on the tool and using advanced machine learning technologies to recognize copyrights. However, since the announcement, the company has never publicly mentioned this tool again. And even if Media Manager is released, experts doubt that the tool will be able to solve all problems.

Adrian Cyhan, an intellectual property lawyer, notes that even large platforms such as YouTube and TikTok have difficulty managing content identification at scale. “Ensuring that content creators and countries comply with all the requirements is an extremely difficult task,” he said. And the founder of the non-profit organization Fairly Trained, Ed Newton-Rex, generally believes that Media Manager will only shift responsibility to the creators themselves.

At the same time, even if Media Manager is launched, it is unlikely to be able to relieve OpenAI from legal liability, experts say. Evan Everist, a copyright specialist, recalled that copyright owners are not required by law to provide notice that their work may not be used, and “the basic principles of copyright law remain the same: you cannot use someone else’s material without permission.”

In the absence of Media Manager, OpenAI has so far implemented filters that prevent verbatim copying of other people’s data, and in lawsuits the company continues to argue that its AI models create a “compilation” rather than plagiarism, citing the principle of “fair use.” Courts may support OpenAI’s position, as happened in the Google Books case, where the court ruled that Google’s copying of millions of books for Google Books, a kind of digital archive, was permissible. However, if the courts find that OpenAI is illegally using copyrighted content, the company will have to rethink its strategy, including the release of Media Manager.