Back in mayOpenAI says it's developing a tool to allow creators to specify how they want to include their creations in their AI training data. But after 7 months, This feature has yet to see the light of day.
The tool, called Media Manager, will “identify copyrighted text, images, audio and video,” OpenAI said at the time, and reflect creators' preferences “across multiple sources.” It was intended to eliminate some parts of the company. The strongest Criticsand potentially protected by OpenAI. Legal challenges to IP.
But people familiar with the matter told TechCrunch that the device is rarely seen as an important launch internally. “I don't think it's a priority,” said one former OpenAI employee. “Honestly, I don't remember anyone doing this job.”
A non-employee coordinating with the company told TechCrunch in December that they had previously discussed the tool with OpenAI. (These people, who are discussing confidential business matters, declined to be identified publicly.)
A member of OpenAI's official team working for media manager Fred von Lohmann transitioned to a part-time consulting role in October. OpenAI PR confirmed Von Lohmann's move in an email to TechCrunch.
OpenAI has yet to provide an update on the progress of Media Manager. The company missed its self-imposed deadline of “2025” to have the tool in place. (To be clear, “in 2025” could be read as including the year 2025, but TechCrunch took OpenAI's language to mean January 1, 2025.)
IP issues
OpenAI uses AI models to learn patterns in data sets to make predictions—for example, A person who bites into a burger will leave a bite mark.. It allows us to learn to some extent how the world works by studying models. ChatGPT Can write convincing emails and essays SoOpenAI's video generator can create fairly realistic footage.
writing The ability to draw on templates to create new movies and other works makes AI incredibly powerful. But it is also regurgitative. When warned in a certain way; countless web pages; Although models trained on videos and images are “publicly available”, It is not intended to be used this way.
For example, Sora can Create clips with TikTok's logo. versus Famous video game characters. The New York Times got ChatGPT to quote its articles verbatim (OpenAI uses the behavior ““hack.“).
This inexplicably upset the creators, whose works were crushed in AI training without their consent. Many are lawyers.
OpenAI allows artists, Fighting class action lawsuits filed by authors. YouTuberscomputer scientists and news organizations; They all claim to be startups who have been trained to do their jobs illegally. The plaintiffs include authors Sarah Silverman and Ta Nehisi-Coates; It includes visual artists and media conglomerates such as The New York Times and Radio-Canada.
OpenAI follows suit. License Compromise together Select partnersBut not all creators see it this way. expressions As attractive.
OpenAI provides several ways for creators to “opt out” of its AI training. Last September, the company was published. A submission form for artists to flag their work for removal from their future training sets. OpenAI has long allowed webmasters to block its web-crawling bots. delete the data. across their domains.
However, creators criticize these methods as artificial and inadequate. written works; There are no separate opt-out mechanisms for videos or audio recordings. The removal form for images is a laborious process requiring a copy of each image to be submitted for removal along with a description.
Media Manager has been commissioned as a complete overhaul and expansion of OpenAI's extraction solutions today.
In a press release, OpenAI said Media Manager will use “the latest machine learning research” to tell creators and content owners what they own (OpenAI). OpenAI says it's working with regulators while developing the tool, adding that it hopes Media Manager will “set a standard in the AI industry.”
OpenAI has not spoken publicly about Media Manager since then.
A spokesperson told TechCrunch in August that the tool was “still in development,” but did not respond to a follow-up request for comment in mid-December.
There's no indication when OpenAI might launch Media Manager — or with what features and capabilities.
Use it sparingly.
Assuming Media Manager does arrive at some point, experts believe it won't assuage creators' concerns — or do much to address legal questions about AI and IP use.
As noted by Adrian Cyhan, IP attorney at Stubbs Alderton & Markiles, Media Manager is an ambitious business. Even big platforms like YouTube and TikTok Struggle together Item ID in dimension. Can OpenAI really do better?
“The warranty is challenging considering the legally required creator protections and the potential for compensation,” Cyhan told TechCrunch. It offers a rapidly evolving and potentially diverse legal landscape, particularly across national and local jurisdictions.”
Ed Newton-Rex, founder of Fairly Trained, a nonprofit that recognizes that AI companies respect creators' rights, believes Media Manager will fairly shift the burden of controlling AI training to creators. They argue that by not using it, they may be giving tacit permission for their work to be used. “Most creators will never hear about it, let alone use it,” he told TechCrunch. “But it will be used to defend the mass exploitation of creative works against the will of the creators.”
Mike Borella, co-chair of MBHB's AI practice group, pointed out that exclusion systems don't always account for changes that might change a job, like a downsampled image. Not all third-party platforms hosting copies of creators' content can handle the same scenario, said Joshua Weigensberg, an IP and media attorney for Pryor Cashman.
“Creators and copyright owners can't control and don't even know where their work appears on the Internet,” says Weigensberg. “Despite the fact that a creator has said they are opting out of training for each AI platform, these companies still make copies of their work and training available on third-party websites and services.”
A media manager for OpenAI would not be particularly advantageous, at least from a judicial perspective. Evan Everist, a partner at Dorsey & Whitney, which specializes in copyright law, said that while OpenAI is working to reduce the exercise of IP-protected content to a judge. The media manager said the company would not be immune from harm. found to have been violated.
“Copyright owners have no obligation to go out and tell others not to infringe on their work before such infringement occurs,” Everist said. “The basics of copyright law still apply — that is, don't copy other people's material without permission. This feature could be more about PR and OpenAI being the ethical user of the content.”
An estimate
Without Media Manager, OpenAI performed filters — Even the imperfect ones — to prevent its models from recompiling training samples. The company continues to file lawsuits. fair use. Defenses, claiming that its models create an artificial transformation;
OpenAI may well succeed in its patent disputes.
Courts may decide that a company's AI has a 'transformative purpose'. Tradition The publishing industry's lawsuit against Google was roughly defined a decade ago. In that case, A court has ruled that Google is not allowed to copy millions of books for Google Books, a type of digital archive.
There is OpenAI. He said. It is publicly announced that it is “impossible” to train competitive AI models without using copyrighted material: permission or not. “Limiting training data to public domain books and images created more than a century ago may provide an interesting experiment, but will not provide AI systems that meet the needs of today's citizens,” the company wrote in a January submission to the UK's House of Lords. .
If the courts ultimately declare OpenAI victorious, the media manager won't serve much of a legal purpose. OpenAI seems willing to make that bet — or rethink its opt-out strategy.