The world’s two biggest artificial intelligence powers—the U.S. and China—are attempting to address intellectual property concerns relating to AI-generated content domestically, as their ability to effectively collaborate on managing AI remains hampered by the dismal state of bilateral relations.
But whether they can find a degree of compromise on AI governance, including IP, will make a big difference to companies. Firms are hoping for the least degree of legislative variation possible between countries.
That’s one reason why Sam Altman went on his AI global tour last summer: As he promoted his version of AI safety to world leaders like Indian Prime Minister Narendra Modi and South Korean President Yoon Suk-yeol, he was also pushing to create a more globally inhabitable environment for his products. He even made a video appearance at a Chinese AI conference, advocating for “global cooperation” and exchanges between Chinese and American researchers.
In the U.S., lawsuits regarding intellectual property and AI-generated content will frequently hinge on the boundaries of “fair use.” AIGC introduces an unprecedented challenge to the current bounds of IP protection, primarily on two levels: input (what data goes in the training model) and output (the creations users make using AIGC-enabled tools).
Stateside, the issue of inputs has become the subject of ongoing court cases: One notable example is The New York Times’ decision to sue OpenAI for copyright infringement on the basis that the AI company used the newspaper’s articles as training data. In China, however, the focus so far has been on outputs. In November, the Beijing Internet Court ruled in favor of protecting AIGC in the first case of its kind. The court ruled the plaintiff met the threshold for “originality” by sufficiently augmenting the AI-generated image, warranting protection.
As Qiheng Chen, a fellow at the Asia Society Policy Institute’s Center for China Analysis, points out in a recent paper, the Beijing’s court’s verdict diverges from decisions made by the U.S. Copyright Office, which has so far opted not to give copyright protections to AIGC. In an interview, Chen explained the Beijing Internet Court’s ruling demonstrates the court is not entirely independent.
“The court may not be immune from high level policy directions,” he said. “The opinions show considerations of the political environment and the preferences of the time.” Amid China’s current economic downturn, the central government’s preferences are geared toward stimulus and industrial development—a posture judges are undoubtedly aware of.
Judge Ge Zhu of the Beijing Internet Court said as much, speaking at a conference late last month hosted by Hong Kong University. Reflecting on the AIGC case, she said, the purpose of copyright law is to encourage creativity, adding that offering copyright protection might help keep creators honest and further develop the market for AIGC. She also noted that failing to protect art generated by AI could create negative incentives for people to conceal how they created the work.
China’s environment of relatively greater interlinkage between courts and regulators (with central regulators as the tone-setters) indicates more centralization in determining the scope of IP rights pertaining to AIGC. Centralization does not guarantee predictability, but Chinese firms are probably closer than their American counterparts to hearing from the central government what protections and curbs they can expect going forward—without having to account for a shift in strategy or direction after, say, a presidential election.
That has implications for the business side of AI. As ongoing court cases, executive orders and regulations continue to come out in both countries, the contours of feasible competition are coming into greater focus.
AI and IP rights has highest stakes for creators, who are looking to use the tools to enhance and speed up their work—but, as Judge Zhu pointed out, they need to know their work will be protected before we can expect mass commercial adoption of AIGC—and all the economic growth that could come with it.
Chen also pointed out the urgency on AIGC and IP lies most clearly with artists. “Creators working in comics, illustration and film need some kind of protection for their creations so they can market and monetize them,” he said, adding that such protections would allow for “more downstream usage of the AI systems, encouraging more innovation.”
China’s advantage, Chen said, is rooted in conditions like “its huge consumer base and user demand, which foster the development of new applications. To some extent, it’s different from the U.S., where the foundational model developers mostly drive downstream development.”
The irony is that the United States’ hands-off stance vis-a-vis technology regulation has always been motivated by supporting innovation. China’s recent AIGC-oriented regulations—like the Interim Measures for AIGC Measurement, released in July—and the Beijing Internet Court’s November ruling have been directly in line with promoting the AI industry.
The two countries have plotted diametrically opposing paths to arrive at conclusions that both appear to rank innovation first—at least for now. That should make cooperation easier: Companies from both countries have international ambitions. But in lieu of substantive global agreements on how to handle the IP AI products both use and help generate, Chinese and American firms—and consumers—are left minding the gaps. They may or may not fill them responsibly.
