Could Chinese AI videos kill Hollywood stars?

Hollywood is outraged by China’s latest AI sensation.

Disney, Netflix, Warner Bros. and several other major movie studios are each threatening to sue ByteDance for copyright infringement after the Chinese company unveiled SeaDance 2.0, a new AI tool that can generate eerily realistic videos of movie actors and cartoon characters.

In the most striking example yet, Irish filmmaker Ruairi Robinson posted a video to X last month showing an epic battle between Brad Pitt and Tom Cruise, in which neither actor played the role.

In response, TikTok’s owner ByteDance pledged to strengthen safeguards to prevent unauthorized use of its intellectual property and canceled Seedance’s international expansion, which was originally planned for last month.

This episode is part of an ongoing discussion about how AI model training data and its output can violate copyright and publicity rights. But it also highlights the growing challenges in regulating emerging technologies and the difficulties that businesses and individuals may face in enforcing their rights, especially across international jurisdictions.

Seedance 2.0 is the keystone of advanced multimodal generative AI capabilities. This is where its biggest rivals, like Meta, lag behind despite tens of billions of dollars in capital investment.

Neil Shah, co-founder of market research firm Counterpoint

“For the West, [Seedance 2.0] Sounds the alarm about the erosion of traditional film production and challenges to cost and profit structures [of the industry]” said Neil Shah, co-founder of market research firm Counterpoint. [the development of] Generative AI technologies make it much harder to set boundaries when it comes to licensing and copyright. ”

The emergence of Seedance 2.0 is a notable example of how far AI models have come.

The prompt “Will Smith eating spaghetti” has long been a popular standard for testing AI’s video generation capabilities. Initial versions of such tools will first become publicly available around 2022; men in black An actor with crooked fingers and a distorted face. Although the results have become more realistic over time, there have always been signs that the video was generated by AI.

Seedance 2.0 produces results that are almost indistinguishable from real video. The character munching on the food sounds like Smith. This demonstrates the tool’s ability to seamlessly combine different media formats, including audio.

“Seedance 2.0 is an important pivot for advanced multimodal generative AI capabilities,” said Shah. “This is where our biggest competitors, like Meta, are lagging behind, despite tens of billions of dollars in capital investment.”

This breakthrough confirms the ability of Chinese companies to innovate and compete despite constraints on financial resources and computing power. According to Artificial Analysis, a San Francisco-based company that evaluates AI video models based on user votes, six of the world’s top 11 video generation models today were developed by Chinese companies, including MiniMax, Alibaba, and short video platform Kuaishou.

ByteDance has an advantage thanks to the hugely popular social media platforms TikTok and its Chinese counterpart Douyin.

“The company has access to vast amounts of visual data, which it can use to train AI models and generate high-quality output,” said Qiang Bai, a former co-founder of Chinese AI company iFlytek who now runs an AI incubator in Japan.

ByteDance also appears to be less concerned about copyright guardrails than its competitors. Rival U.S. tools are working to block the copying of copyrighted material or enter into licensing agreements with intellectual property owners. Google’s Nano Banana, for example, rejects requests to create videos that could harm “the interests of third-party content providers.” OpenAI signed a deal with Disney in December that allows users to legally generate content featuring Disney characters.

In contrast, Seedance 2.0 allows users to have Spider-Man fight the video game character Jinx and create alternate endings. titanicproduces a viral music video with Kanye West singing in Chinese. Bai says Seedance 2.0 has become popular precisely because it allows users to recreate familiar faces and scenes.

Two major issues arise in such a use case. First, if a company uses copyrighted material to train an AI model, is that a so-called “fair use” of the content? Second, whether such output infringes copyright or publicity rights. Both issues are currently being heard in courts in both the United States and China.

AI video generated using ByteDance’s Seedance 2.0 software. Source: Aleena Amir, Martin, Shuyang Sihu

In the United States, Disney, Universal, DreamWorks, and most recently Warner Bros. are suing San Francisco-based AI startup Midjourney for copying images of Darth Vader, the Minions, Superman, and others without permission, and the case is currently in the discovery phase. Similar lawsuits have been brought against AI companies by artists and labor unions. In progress.

Suing foreign companies is even trickier, partly because of the red tape.

In September, Disney and other movie studios filed a lawsuit in the United States accusing Chinese AI company MiniMax of “willful and brazen” copyright infringement and seeking up to $150,000 in damages for each infringing work. Users of MiniMax’s Hailuo AI were able to easily generate iconic characters from the Joker to Wonder Woman using simple text prompts, according to the complaint. The company also said it used such material to promote its tools, promoting itself as “Hollywood studios in your pocket.”

However, it took nearly six months just to submit the complaint to Minimax. This was partly because it had to be done through procedures set out in the Hague Convention, a set of international treaties that help resolve cross-border legal disputes. MiniMax’s U.S. attorney accepted the charges in late February.

MiniMax dismissed the claims as “unfounded” in its recent initial public offering prospectus, which raised $619 million in its Hong Kong trading debut in January.

“It will likely be difficult for U.S. IP owners to obtain meaningful relief against Chinese companies that do not have significant assets in the U.S.,” said Angela Chan, a law professor at the University of Southern California. “Even if the plaintiff were successful on the merits, it would be difficult to enforce the judgment in an extraterritorial manner.”

China doesn’t have the same tradition as Hollywood, so it’s less hesitant to embrace something new. Content generation could be a silver bullet for Chinese AI companies.

Qiang Bai, former co-founder of Chinese AI company iFlytek, now runs an AI incubator in Japan

In a landmark case in 2024, a Hangzhou court ruled that a Chinese AI platform’s use of images of the Japanese cartoon character “Ultraman” to train AI models is considered “reasonable” under China’s copyright law. “This decision is widely seen as representing a more permissive approach to the use of copyrighted material as input for AI training,” Zhang said.

But in a separate case, a Guangzhou court ruled that a Chinese company whose website could use text prompts to generate images substantially similar to Ultraman had infringed copyright. The company was ordered to pay compensation of 10,000 yuan ($1,450), which is part of the amount demanded by the copyright owner.

He Tianxiang, an associate professor of law at City University of Hong Kong, said movie studios could file strong lawsuits against ByteDance under Chinese law over output such as AI videos that incorporate real or fictional characters from copyrighted films. “From what we’ve seen, we think this is a clear violation of copyright law,” he says.

The challenge lies in cross-border litigation, which can be as much a political as a legal process.

“This is very complex because of its international nature,” says Los Angeles entertainment and technology attorney Jonathan Handel. “I don’t know anyone in the American corporate system who really trusts the Chinese legal system to justify their rights.”

A bigger question is the impact that AI technology will have on the future of the entertainment industry.

Handel expects it will take another two to three years for AI technology to evolve to the point where it can produce feature films good enough to be shown in theaters. But “the realistic nature of AI video has almost everyone in Hollywood very concerned about the future of their jobs and filmmaking,” he says.

The backlash against Seedance 2.0 is just the latest example of the U.S. entertainment industry’s anxiety about AI. Actors and writers have gone on strike to protest its unregulated use, while producers of films such as civil war and brutalist AI is being attacked for using it to enhance audio and create movie posters.

By comparison, China’s creative industries have embraced AI technology, seeing it as a path to making money.

Renowned filmmaker Jia Zhangke collaborated with ByteDance to release a short film last month in which he talks to an AI avatar. “You provide the ideas and I provide the computing power,” the Avatar says. In the same month, Liu Cixin, the author of the world-famous science fiction blockbuster novel, three body problemSo, we produced a short drama made with visuals generated by AI. The ability to use AI tools has become a must for filmmakers in this era, the director said in an interview.

AI studios are springing up across China, producing reels, dramas and ads, while traditional film companies are cutting back on projects involving human actors. Chinese market research firm iiMedia estimates that the country’s AI-generated content market is expected to double this year to reach $23 billion.

“China doesn’t have the same tradition as Hollywood, so we’re not too hesitant to introduce new things,” Bai says. “Content generation could be a silver bullet for Chinese AI companies.”

---