A group of Japanese companies, including Studio Ghibli, have written a request to OpenAI, asking it to stop training Sora 2 on copyrighted work. Earlier this year, using ChatGPT to generate Studio Ghibli-style images became something of an online craze, with even OpenAI boss Sam Altman changing his profile pic to a self portrait made ... Studio Ghibli among a group of Japanese companies demanding OpenAI to stop stealing their work
A group of Japanese companies, including Studio Ghibli, have written a request to OpenAI, asking it to stop training Sora 2 on copyrighted work.
Earlier this year, using ChatGPT to generate Studio Ghibli-style images became something of an online craze, with even OpenAI boss Sam Altman changing his profile pic to a self portrait made in the style of the much-loved animation house.
Since then, Sora 2, OpenAI’s combination of AI video generator and TikTok-esque social platform, has raised more questions over copyright laws – and whether they even mean much if someone from Ipswich can type in a brief prompt and generate a clip of Pikachu being roasted on a barbecue.
A group of Japanese companies has now written to OpenAI, demanding that its “content is not used for machine learning without their permission.”
That letter was written by CODA – the Content Overseas Distribution Association – an organisation set up in 2002 to, among other things, prevent Japanese work from being pirated in foreign territories. Among CODA’s members are some of Japan’s biggest names in videogames, film and television, including Studio Ghibli, Toei Animation, Toho, Bandai Namco and more besides.
As first reported by Automaton, CODA wrote its open letter on the 28th October in response to the launch of Sora 2 on the previous month. The organisation writes that, “a large portion of content produced by Sora 2 closely resembles Japanese content or images,” and that “this is the result of using Japanese content as machine learning data.”
The letter then argues that the process of feeding this content into Sora 2 for later use in the platform’s output may “constitute copyright infringement,” and demands that OpenAI stops using Japanese companies’ work without their permission.
CODA also politely yet firmly responds to OpenAI’s suggestion in September that any individual or company that didn’t want their work to be taken by Sora 2 would have to “opt out” – in other words, they’d have to write to OpenAI and request that their material be removed from the platform’s training data.
“It has been stated that Sora 2 responds through an opt-out system based on requests from copyright holders,” CODA writes. “However, under Japan’s copyright system, prior permission is generally required for the use of copyrighted works, and there is no system allowing one to avoid liability for infringement through subsequent objections.”
Altman later wrote in a blog post that his company would provide more “granular control” to rightsholders, and as it became clear how easy it was to generate weird and potentially copyright-infringing clips using Sora 2, a number of ‘guardrails’ were put in place within days of its launch.
Still, the question of what data has been used to ‘train’ Sora 2 and platforms like it remain, and a number of lawsuits could decide exactly what the limits are when it comes to the question of generative AI and copyright.
Suggested product
Issue #60 Magazine Bundle (3x Covers: A Scanner Darkly Magazine + 1991 + Roger Deakins)
£14.99
A group of authors, including Game Of Thrones author George RR Martin, are collectively suing OpenAI over the company’s (alleged) use of their writing. Disney and Universal have clubbed together to sue the AI firm Midjourney, calling it a “bottomless pit of plagiarism.”
When Sora 2 launched, Sam Altman excitedly said that the firms he spoke to were “excited” about his new computer program, and said they regarded its output as “interactive fan fiction.” It’s probably fair to say that the companies represented by CODA don’t feel this way.
