Website Owner's Home (ChinaZ.com) June 21 News: The Beijing Internet Court recently heard the city's first two cases of "AI face-swapping" software infringement. The plaintiffs, Liao and Wu, are models in Chinese-style short videos. They accused the operator of a face-swapping app of using their videos to create face-swap templates without authorization and offering them for paid use in the app, infringing on their portrait rights and personal information rights.
The court found that although the defendant used the plaintiffs' videos for deep synthesis technology to replace the faces in the videos, this did not constitute an infringement of the plaintiffs' portrait rights, as the replaced videos could no longer identify the plaintiffs. However, the court also determined that the defendant's actions infringed on the plaintiffs' personal information rights.
The court pointed out that the facial features and other individualized information in the plaintiffs' videos are personal information, and the defendant's use of face-swapping technology to process this information constitutes personal information processing. The defendant's unauthorized acquisition and commercial use of the plaintiffs' personal information constituted an infringement.
Ultimately, the court ordered the defendant to apologize to the plaintiffs and compensate for their mental and economic damages. The case is still within the appeal period, and the first-instance judgment has not yet taken effect.
Judgment:
The defendant did use the plaintiffs' videos, but did not constitute an infringement of their portrait rights
After the court hearing, it was found that the defendant did not submit evidence proving the source of the template videos, and considering the consistency of the characters' makeup, hairstyle, clothing, actions, lighting, and camera switching in the template videos with those in the plaintiffs' videos, it can be concluded that the defendant used the plaintiffs' videos, replaced the faces with those of others through deep synthesis technology, and then uploaded them to the app as templates for user use. However, this did not infringe on the plaintiffs' portrait rights.
Firstly, the face-swapped template videos do not have the recognizability of portraits. Recognizability emphasizes that the essence of a portrait is to point to a specific person, and the portrait reproduced by technical means should be able to be recognized by the public as the image of a certain person. Although with the development of the times and technology, the scope of portrait rights protection is not limited to the face, it should still comply with the legal stipulation that "reflects the external image of a specific natural person that can be identified", and can form a one-to-one correspondence with the specific natural person. In this case, the characters' faces in the videos not only were removed but also replaced, essentially replacing the core part of the video that was recognizable with the face of another recognizable person, eliminating or even destroying the recognizability of the plaintiffs' videos, and the public could directly identify the characters in the face-swapped template videos rather than the plaintiffs, unable to form a one-to-one correspondence with the plaintiffs.
Secondly, the defendant did not commit the legally defined acts of infringing on the plaintiffs' portrait rights. According to the Civil Code, acts of infringing on portrait rights include unauthorized production, use, or public disclosure of a person's portrait, defamation or contamination of the portrait, or the use of information technology to forge another person's portrait. In this case, the defendant did not produce videos containing the plaintiffs' portraits; although the defendant used the plaintiffs' videos, it was not a use of the plaintiffs' portraits but a replacement of the face that could identify the plaintiffs, thereby removing the recognizability of the portrait and then utilizing the non-personal elements of the video, such as makeup, clothing, hairstyle, lighting, and camera switching, to obtain property interests; furthermore, the defendant did not defame or contaminate the plaintiffs' portraits; and the defendant's actions also did not constitute forging the plaintiffs' portraits.
Therefore, the defendant's actions do not fall under the legal definition of infringing on the plaintiffs' portrait rights and did not infringe on the personal and property interests attached to the plaintiffs' portraits.
The defendant's actions constitute an infringement on the plaintiffs' personal information rights
First, the plaintiffs' videos contain personal information including their facial information. The plaintiffs' videos dynamically present their facial features and other individualized characteristics. Based on digital technology, these personal characteristics can be presented in data form, in line with the definition of "information related to an identified or identifiable natural person" in the Personal Information Protection Law of the People's Republic of China.
Second, the defendant engaged in the processing of the plaintiffs' personal information. Firstly, the defendant should be the subject responsible for the processing of personal information. Even if the defendant actually used the technical services of a third-party company, the third-party company is only the entrusted technical service provider, and the defendant is the principal who decides the methods and scope of information processing and should bear the responsibility for the processing of personal information. Secondly, the face-swapping behavior constitutes personal information processing. The defendant first needed to collect the plaintiffs' videos containing their facial information, replace the plaintiffs' faces in the videos with the faces in the photos provided by the defendant, a process that used facial recognition technology to detect key facial points, and then fuse the facial features of the provided image into the specific character in the template image, resulting in an image that combines the facial features of both the specified image and the template image. This synthesis process is not a simple replacement but requires the fusion of the features of the new static image with the facial features, expressions, etc., of the original video through an algorithm, making the replaced template video appear natural and smooth. This process involves the collection, use, and analysis of the plaintiffs' personal information, so the formation of the face-swapped template video through face-swapping constitutes the processing of the plaintiffs' personal information.
Third, the defendant's actions infringed on the plaintiffs' personal information rights. The automated processing of personal information often has hidden characteristics, so the law empowers individuals with the right to know and the right to decide regarding the processing of their personal information to prevent risks such as leakage and misuse. Although the plaintiffs' videos were already public, the account description stated "not authorized for any paid software," and it should not be assumed that the plaintiffs consented to the processing of their facial information. In addition, the defendant obtained the videos containing the plaintiffs' facial information, analyzed and modified them using deep synthesis, an emerging technology, and then commercially exploited them, which could have a significant impact on the plaintiffs' personal rights and should be conducted with the plaintiffs' consent. The defendant has no evidence to prove that they obtained the plaintiffs' consent, so it constitutes an infringement on the plaintiffs' personal information rights.