According to the Star Daily on January 6 (Reporter Huang Xinyi), iQIYI has formally filed a lawsuit with the People's Court of Xuhui District in Shanghai, accusing the domestic AI startup MiniMax of infringing its copyright in the AI model training and content generation process.

In response, iQIYI stated to the Star Daily that the incident is still under legal proceedings and no further information can be disclosed. MiniMax has yet to respond to this matter.

It is reported that MiniMax is accused of using copyrighted materials owned by iQIYI for model training without authorization, resulting in generated content that constitutes an infringement of iQIYI's copyright. iQIYI demands that MiniMax immediately cease this infringing behavior and seeks compensation of approximately 0.1 million yuan.

On MiniMax's subsidiary Hai Luo AI creative video platform, users can generate videos by uploading images and text.

The Hai Luo AI user agreement states that users must ensure that the content they upload while using the Hai Luo AI software and related services is either original or legitimately authorized, and does not infringe on anyone's intellectual property rights, reputation rights, name rights, privacy rights, commercial secrets, or any other legal rights. Otherwise, the user will bear the risk and responsibility for any resulting infringement, and the corresponding disputes will be the user's responsibility to resolve, along with compensating Hai Luo AI for all direct and indirect losses suffered.

The formation and improvement of AIGC models depend on extensive data training, and the data used for training often contains content protected by copyright law, leading to numerous relevant copyright disputes both domestically and internationally.

In June 2024, a domestic AIGC creator named DynamicWang accused 360 of stealing images generated by his AI drawing model during the launch of its new AI product and demonstrated the product's "partial redrawing" feature during the event.

On November 29, 2023, four domestic painting creators announced on Weibo that they have filed a lawsuit against the parent company of Xiaohongshu and the parent company of Xiaohongshu Trik Software in court, claiming that Trik used the artists' original works as training data without authorization, thus generating images that are highly similar to the originals, "infringing on the creators' legal rights."

On January 23, 2023, three USA comic artists initiated a class action lawsuit against three AIGC commercial application companies, including Stability AI, in the Northern District Court of California, accusing the Stable Diffusion model developed by Stability AI and the respective paid AI image generation tools launched by the three defendants, based on the aforementioned model, of copyright infringement.

There have already been related judicial precedents in the domestic copyright disputes concerning AI images. In January 2024, the first case of copyright infringement involving AI-generated images in China became effective. After the image generated using the Stable Diffusion model was infringed, the plaintiff filed a lawsuit in the Peking Internet Court. This case went through five trials, and the court ultimately determined that the plaintiff's image possesses "originality," meets the definition of a work, qualifies as a work of fine arts, and is protected by copyright law.

However, regarding the issue of unauthorized training data, related cases domestically and internationally are still in the judicial process, and no effective judgment has been made. Lawyer Xiong Dingzhong from Peking Qinglu Law Firm told the reporter from the Star Daily that the rights of AIGC-generated products (if any) have basically formed a consensus in the industry, namely that AIGC companies do not assert rights, and the issue is more about balancing the rights and obligations between users. However, as of now, there is still considerable controversy in the industry regarding the issue of training data, which is also the most concerning issue for the industry. The core reason mainly lies in the fact that, for the AIGC industry, its entire economic revenue can come from users' paid usage behaviors. Currently, the mainstream AIGC products in the world all use this model, but whether obtaining the consent of the original rights holders (whether it is personality rights such as portrait rights, privacy rights, personal information rights, or copyright) is still an unresolved issue. How to balance protecting the legal rights and interests of rights holders and promoting the development of the AI industry is a matter of close concern for global legislation and judicial agencies.

Xiong Dingzhong believes that granting full respect to the original rights holders not only reflects recognition and respect for human intelligence but also ensures that human creators continue to produce high-quality training resources for AI. Even from the perspective of the most ardent supporters of AI, a predatory approach to the issue of training data should not be adopted.