Back in May, OpenAI announced that it was developing a tool that will allow creators to find out how their works needs to be included or excluded from AI training data. But seven months later, this feature still hasn't seen the sunshine of day.
The tool, called Media Manager, would “discover copyrighted text, images, audio and video files,” OpenAI said on the time, to reflect creators' preferences “across multiple sources.” It was intended to fend off among the company's harshest critics and potentially protect OpenAI from legal challenges related to mental property.
But people familiar tell TechCrunch that the tool was rarely seen as a serious launch internally. “I don’t think it was a priority,” said a former OpenAI worker. “To be honest, I don’t remember anyone working on it.”
A non-employee who coordinates work with the corporate told TechCrunch in December that the tool had been discussed with OpenAI previously, but there had been no recent updates. (These people declined to be identified publicly discussing confidential business matters.)
And a member of OpenAI's legal team who worked on the media manager, Fred von Lohmann, moved to a part-time advisory role in October. OpenAI PR confirmed Von Lohmann's move to TechCrunch via email.
OpenAI has yet to supply an update on Media Manager's progress and the corporate has missed its self-imposed deadline for launching the tool by 2025.
IP problems
AI models like OpenAI learn patterns in data sets to make predictions – for instance this Anyone who bites right into a burger leaves a bite mark. This allows models to learn, to some extent, how the world works through remark. ChatGPT can write compelling emails and essays, while Sora, OpenAI's video generator, can create relatively realistic footage.
The ability to attract on examples from texts, movies, and more to generate recent works makes AI incredibly powerful. But it's also belching. When asked to achieve this in a certain way, models – most of that are trained on countless web pages, videos and pictures – create near-copies of that data, which, while “publicly available,” is just not intended for such use .
Sora can, for instance Generate clips with the TikTok logo and popular video game characters. The New York Times got ChatGPT to cite their articles verbatim (OpenAI made “chop“).
This has understandably angered creators whose works have come under AI training without their permission. Many have turned to lawyers.
OpenAI is fighting class-action lawsuits from artists, writers, YouTubers, computer scientists and news organizations, all of whom claim the startup worked on their works illegally. Plaintiffs include authors Sarah Silverman and Ta Nehisi-Coates, visual artists and media corporations comparable to The New York Times and Radio-Canada.
OpenAI has licensing agreements with select partners, but not all developers see this Conditions equally attractive.
OpenAI offers developers several ad hoc ways to opt out of its AI training. Last September, the corporate introduced a submission form that permits artists to mark their work for removal from future training sets. And OpenAI has long allowed webmasters to dam its web crawling bots from collecting data across their domains.
But developers have criticized these methods as arbitrary and inadequate. There aren’t any specific opt-out mechanisms for written works, videos or audio recordings. And the image opt-out form requires submitting a replica of every image to be removed together with an outline, a laborious process.
Media Manager was launched today as an entire overhaul – and expansion – of OpenAI's opt-out solutions.
In the May announcement, OpenAI said Media Manager would leverage “cutting-edge machine learning research” to present creators and content owners the power to “tell OpenAI what they own.” OpenAI, which claimed it was working with regulators to develop the tool, said it hoped Media Manager would “set a normal across the AI industry.”
OpenAI has never publicly mentioned Media Manager since then.
A spokesperson told TechCrunch that the tool was “still in development” in August, but didn’t reply to a follow-up request for comment in mid-December.
OpenAI has given no indication of when Media Manager might launch – and even what features and capabilities it’d launch with.
Fair Use
Assuming Media Manager eventually involves market, experts aren't convinced it’s going to assuage creators' concerns – or do much to resolve the legal issues surrounding using AI and mental property.
Adrian Cyhan, IP lawyer at Stubbs Alderton & Markiles, noted that Media Manager as described is an ambitious undertaking. Even platforms as big as YouTube and TikTok Battle with Content ID at scale. Could OpenAI really do it higher?
“Ensuring compliance with statutory copyright protections and potential compensation requirements into account is difficult,” Cyhan told TechCrunch, “particularly given the rapidly evolving and potentially different legal landscape across national and native jurisdictions.”
Ed Newton-Rex, the founding father of Fairly Trained, a nonprofit that certifies that AI corporations respect creators' rights, believes Media Manager would unfairly shift the burden of controlling AI training onto creators; If they don't use it, they could be giving tacit permission to make use of their works. “Most developers won’t ever hear of it, let alone use it,” he told TechCrunch. “But it continues to be used to defend the mass exploitation of creative works against the need of the creators.”
Mike Borella, co-chair of the MBHB's AI practice group, identified that opt-out systems don’t at all times take into consideration transformations that might be made to a plant, comparable to: B. a downsampled image. They also may not take into consideration the all-too-common scenario of third-party platforms hosting copies of creators' content, added Joshua Weigensberg, an IP and media attorney at Pryor Cashman.
“Creators and copyright holders haven’t any control and infrequently don’t even know where their works appear on the Internet,” said Weigensberg. “Even if a creator tells each individual AI platform that they’re opting out of coaching, those corporations should still go ahead and train using copies of their works which can be available on third-party web sites and services.”
At least from a legal perspective, Media Manager may not even be particularly useful to OpenAI. Evan Everist, a partner at Dorsey & Whitney who makes a speciality of copyright law, said that while OpenAI could use the tool to point out a judge that it’s softening its training on IP-protected content, media managers are unlikely to guard the corporate from harm If this were the case, it might be determined that a violation had occurred.
“Copyright holders haven’t any obligation to preemptively tell others to not infringe their works before that infringement occurs,” Everist said. “The basics of copyright law still apply – that’s, you usually are not allowed to adopt and replica third-party content without permission. This feature could also be more about PR and positioning OpenAI as an ethical user of content.”
A reckoning
In the absence of Media Manager, OpenAI has implemented filters – albeit imperfect – to forestall his models from reproducing training examples. And within the lawsuits the corporate is fighting, it continues to argue that fair use protections exist and claims that its models create transformative, not plagiaristic, works.
OpenAI could thoroughly prevail in its copyright disputes.
The courts could resolve that the corporate's AI has a “transformative purpose,” following the precedent set within the publishing industry's lawsuit against Google a few decade ago. In this case, a court ruled that Google's copying of tens of millions of books was legal for Google Books, a style of digital archive.
OpenAI has said publicly stated that it’s “inconceivable” to coach competitive AI models without using copyrighted material – whether authorized or not. “Limiting training data to public domain books and drawings created greater than a century ago could make for an interesting experiment, but wouldn’t deliver AI systems that meet the needs of today's residents,” the corporate wrote in a press release Submitted to the House of Lords in January.
If courts ultimately declare OpenAI victorious, Media Manager wouldn't serve much of a legal purpose. OpenAI appears able to make that bet – or rethink its opt-out strategy.
