- The Abacus
- Posts
- China Protects AI-Generated Image Copyright
China Protects AI-Generated Image Copyright
What are Beijing's criteria and how do China and the US differ in their approach to AI artwork copyrights
With AI getting big, copyright issues are popping up more and more. It's turning into a real head-scratcher, blurring the lines between human and machine creativity. Over in the U.S., the Copyright Office just said 'no' for the fourth time to AI-created stuff. Meanwhile, China seems to be taking a different route with its first AI copyright cases, kind of encouraging people to use AI for creative stuff.
This week, I've written two newsletters to really dig deep into this issue. Today's newsletter will cover:
What are Beijing's criteria and how do China and the US differ in their approach to AI artwork copyrights
What signal are Chinese authorities sending about AI with their latest ruling on copyrights
Enjoy it!
I
Let’s look back at the case.
In March 2023, Mr. Li began experimenting with Stable Diffusion, a text-to-image AI system from the U.S. He created an illustration of a girl after inputting hundreds of prompt words and constantly adjusting AI settings and variables, then shared the final one on Xiaohongshu.
One blogger used the image without permission and removed Mr. Li's watermark for her blog. Mr. Li viewed this as an infringement of his rights, sparking a debate about the ownership of AI-generated art.
On November 27, the case reached a conclusion. The judge granted the 'work' nature of AI-generated images and the “human authorship” status of the user, awarding Mr. Li the copyrights to this AI-generated image.
Meanwhile in the U.S., the artwork 'SURYAST,' which was created from an original photo using the AI software RAGHAV in a Van Gogh-like style, faced copyright challenges. The creators twice requested reconsideration for copyright protection, but these efforts were denied. This incident marks the U.S. Copyright Office's fourth consecutive refusal to register AI-generated works.
In the Chinese case, the AI-generated image was classified as “fine art work,” a stark contrast to the U.S. “SURYAST” case, labeled as a “derivative work.”
The Beijing Internet Court ruled that the AI-generated image met the criteria of “works” under the Copyright Law, as it displayed originality and intellectual creation in the artistic domain.
Both the U.S. Copyright Office and China's courts concur that the author of a “work” must be a human being, not an AI like large language models (LLMs). Globally, the key determinant of a work's copyright eligibility hinges on the extent of human involvement. This creates a blurred boundary with no clear definition in current laws.
When the responsibility for interpretation lies with governmental authorities, it becomes a critical moment, I believe, to clearly understand the government's stance on AI.
In the previous U.S. case of Théâtre D’opéra Spatial, the author used over 600 text prompts to create the work and even won an award in an art competition. Despite this, copyright was not recognized. The U.S. Copyright Office seems to believe that users cannot consistently predict the outcome of AI-generated content, suggesting that the AI system exercises significant control over the final product, leading to the denial of human authorship.
Cui GuoBin, a law professor at Tsinghua University, believes that “this approach does not align with reality.”
Zhu Ge, the judge in the Chinese case, stressed the need for case-by-case judgments for AI-generated content, centering on the degree of human creative control. On the court website, she outlined the factors she considers when making a ruling:
"Originality is determined by courts around the world through individual case judgments. This involves academic discussions, learning from legal practices in other countries, and importantly, judges weighing various interests. They consider the work's nature, its creative possibilities, relevant industry policies, and what the public needs. Each judge strives to interpret these aspects using their country's legal standards and language."
Judge Zhu notes that as technology advances, more people will use AI for creating works, a shift from the manual ways assumed by copyright law. She highlights that choices in AI art, like prompt words and artistic decisions, show intellectual creation input. She believes that “Traditional theories need updating in response to these changes.”
II
So, when China protects AI-generated image copyright, what are they really protecting?
Hu Ling, an associate professor at Peking University Law School, argues that this landmark AI-related copyright case safeguards the user's investment in learning and utilizing the software, as well as the intellectual effort involved in adjusting and designing to create better works.
Unlike DALL·E, where users simply input prompt words to generate an image, Stable Diffusion requires a certain level of skill from its users, presenting a higher barrier to entry. Without legal protection, this lack of accessibility could diminish enthusiasm for learning to use such AI software.
This is a nuanced issue. It's important to remember that people in China cannot directly access overseas AI software like ChatGPT, Midjourney, and others by simply searching online. They must navigate around the internet restrictions imposed by Chinese authorities to acquaint themselves with the latest AI technologies.
In this context, the judiciary's more laxer approach is encouraging more and more people to explore content creation using AI software platforms. This also helps to enhance the commercial value of AI products and services, even though there is an awareness of internet limitations on certain services. And could benefit big tech companies, in the long run.
Industry
ByteDance is negotiating with several firms, including Tencent, to sell its gaming assets. (Josh Ye / Reuters)
Nividia to launch China-focused AI chip in Q2 2024. (Yelin Mo and Fanny Potkin / Reuters)
Microsoft launches AI Odyssey to train 100,000 Indian developers in advanced AI technologies and tools. (Microsoft)
Chinese firms are adapting Nvidia gaming chips for AI use. (Qianer Liu / FT)
Microsoft has faced questions over whether operating an advanced research lab in Beijing is politically tenable.( Karen Weise, Cade Metz and David McCabe / New york Times)
Policy
Tencent Games limits minors to 16 hours of gaming during the 2024 winter vacation.
China's state-backed institution can now identify AirDrop message senders, aiding Beijing's push against undesirable content. (Bloomberg)
U.S.-developed RISC-V chip technology, crucial for China's goals, is under Washington's scrutiny for potential restrictions. (Don Clark and Ana Swanson / New York Times)
Bipartisan US House lawmakers urge the Biden administration for firmer measures against China's rising dominance in older-generation chip manufacturing. (Asa Fitch / WSJ)
China's National Data Administration announces a three-year action plan aiming for a 20% annual growth in the data sector to bolster the economy. (Xinmei Shen / SCMP)
Global elections are increasingly challenged by disinformation, fueled by extremism, AI advancements, and insufficient content moderation. (Tiffany Hsu, Stuart A. Thompson and Steven Lee Myers / The New York Times)
Reply