Opinion | Research on legal issues of video-generated artificial intelligence Sora

Scan the QR code to subscribe to “China Information Security”

Mailing code 2-786

Subscription hotline: 010-82341063

Text |Beijing Normal University Law School Song Jin

Recently, the famous American technology company OpenAI launched the world's first video generation model Sora, which once again attracted great attention from the world. Not long ago, the generative artificial intelligence ChatGPT, also launched by OpenAI, has aroused heated discussions due to its powerful performance and ability to generate fakes that resemble real ones.

The previous ChatGPT can generate answers that are convincing enough based on the chat content with users and based on previously input training data. It can also be used for text production, email writing and even paper creation. Its outstanding performance not only demonstrates the huge potential of generative artificial intelligence in the work field, but also raises concerns about academic ethics, false information, information security, etc.

The Sora launched this time belongs to the generative artificial intelligence AIGC (Artificial Intelligence Generated Content), OpenAI calls it a video generation model in a report on its official website (video generation models), as the name suggests, generative artificial intelligence is a technology that uses algorithmic technology to input a large amount of existing data for learning and training to create new original content. Unlike ChatGPT, which focuses on producing text content, the Sora launched this time uses input Massive video data enables the production of video content. The report states that Sora can generate videos and images of varying durations, aspect ratios and resolutions, up to one minute of high-definition video.

In fact, Sora is not the first time that humans have used artificial intelligence in the field of video production. Previously, artificial intelligence has been widely used in video editing, such as automatically matching background music, generating titles, etc., but in this process artificial intelligence mostly plays a role. The “auxiliary” role still requires more manual operations in specific production. This time Sora is more of a “creator”. In layman's terms, its creation method is a kind of “wensheng video”. Users only need to use text to describe the content they want without providing any image materials, and Sora can Produce high-definition videos of up to one minute on request.

Before this, there were also videos created through generative artificial intelligence on the Internet, but most of them were created through the more mature “Wen Shengtu” artificial intelligence, which indirectly realized video creation by producing multiple pictures for continuous playback, and its spatial relationship, proportional projection, etc. There are many flaws, and it is impossible to independently generate sounds to match them. This time, Sora has established a “world model”. The objects in the picture completely comply with the laws of real physics. The relationship between light, shadow, space, etc. in the picture is normal, which can provide a real visual perception. It is also matched with real-life objects that are consistent with the picture. Auditory experience.

At present, the discussion about Sora is still divided into two aspects. On the one hand, people are excited about its broad prospects for application in the future media field. Some commentators praise it as “one of the biggest leaps in technology since the emergence of the Internet.” “; On the other hand, it is full of hidden worries about the potential risks and challenges it brings, including false information and industry impact.

So, in the legal field, what potential legal issues will the video-generating artificial intelligence represented by Sora face?

Risks of abuse of deepfake technology

Previously, there have been many cases on the Internet where artificial intelligence “face-changing” technology was used to forge personal indecent photos and videos for dissemination to damage reputation and extortion, and cases where artificial intelligence was used to create false videos to pretend to be victims' relatives and friends to defraud money. , causing serious damage to the victim’s privacy, property rights and other rights.

In fact, due to the limitations of previous artificial intelligence technology, the videos generated by such crimes still have certain limitations in terms of accuracy and have many loopholes. It is still possible to see through them if they are carefully identified, and the video “face-changing” technology still requires the producer to provide as footage The main part, the “matrix” video, not only increases the possibility of the victim seeing through the crime, but also makes the crime operation more difficult. In addition, since a certain computer foundation is still required, the implementation of “face-changing” is still a high-tech crime to a certain extent and has a high technical threshold.

The emergence of Sora, the video generation model, makes it possible to produce fake videos with higher accuracy and greater realism. At the same time, the natural chat-style human-computer interaction mode also allows people with almost no knowledge of artificial intelligence technology to Being able to easily use this technology to generate the required videos further reduces the technical difficulty of such crimes and makes them difficult to prevent.

In addition, Sora also has the ability to expand videos forward or backward in time, convert input video environments, and connect multiple input videos to achieve “seamless connection” and “seamless editing” of videos, integrating AI creation with real-life The existing videos are connected into one, making it even more difficult to distinguish between true and false. Specific to the judicial field, if the parties use this technology to tamper with and splice corresponding video evidence, it may pose severe challenges to the judicial authorities' authenticity inspection of the evidence.

Copyright issues in videos generated by artificial intelligence

Since the advent of generative artificial intelligence, discussions on whether copyright can be granted to content generated by artificial intelligence and who should be granted copyright have been long-standing controversial topics in both theory and practice.

Those who hold the “artificial intelligence tool theory” believe that artificial intelligence does not have legal subject qualifications, that is, it does not have the status of “author”, so its creations also lack “originality” and cannot be protected by copyright law. The wide application and potential huge economic benefits of generative artificial intelligence in reality mean that if you insist on denying its copyright protection, it may damage users' enthusiasm for using the technology and the market's motivation to invest in research and development of this technology, and hinder the progress of the technology.

The lagging nature of the law determines that it cannot completely “predict the future” and cannot stipulate in advance one by one new technologies that are changing rapidly in society. For such content that has not been clearly stipulated in the law, judicial practice will test the judge's understanding of its nature, and in reality, conflicting and contradictory judgments may result.

Even if copyright is granted to works generated by artificial intelligence, the division of ownership is still an urgent legal issue that needs to be resolved.

If artificial intelligence itself is fictionalized as the “author” and the copyright belongs to the artificial intelligence itself, this approach may not only cause many problems in terms of ethics and morality, but also with the current technical level, artificial intelligence still cannot be as full as a natural person. Carry out activities such as maintenance and use of relevant rights.

Therefore, copyright still needs to point to a concrete natural person. In a common creation scenario, it will first involve two subjects – the user who issued the creation instruction and the developer of the artificial intelligence model. When it comes to the degree of originality between users and developers, generative artificial intelligence's way of issuing commands like daily chats not only simplifies the difficulty of use, but also weakens the user's originality; while artificial intelligence The “technical black box” created by inputting massive amounts of data for independent training and improvement has also questioned the originality of developers. There is a theoretical dispute over who should be given the copyright to the content generated by artificial intelligence.

In addition, the emergence of Sora has the most direct impact on the film and television industry, which will also bring new subjects of dispute over the issue of copyright ownership.In the field of film and television, excellent shots and beautiful pictures are of course indispensable for successful films, but the popularity and social influence of the actors in the film are also important factors. Video artificial intelligence has the ability to generate videos with real actors as characters. If a film and television work is produced from this, can actors who did not actually participate in the creation but provide portraits enjoy the corresponding legal “performer's rights” Or some new type of rights, this issue still needs to be further discussed and determined at the level of legal norms.

Remaining remarks

The emergence of every emerging thing in social life may trigger a corresponding series of legal issues. The emergence of video-generating artificial intelligence represented by Sora, while showing great potential in the cultural industry and other fields, will also face many legal issues including but not limited to the risks of deep forgery technology and disputes over video copyright ownership.

Through the law as a regulator of social relations, we cannot avoid responding to new things. We can make adjustments and regulations according to actual conditions, such as improving relevant laws and regulations in the fields of intellectual property and other fields, so as to achieve the unity of promoting scientific and technological progress and maintaining social order.

At the same time, comprehensive governance should be achieved by establishing industry self-discipline conventions, shaping technical ethics standards, and enhancing user-related awareness to keep the development of AI technology on a healthy track.

(Source: Rule of Law Daily)

Share network security knowledge and strengthen network security awareness

Welcome to follow the official Douyin account of “China Information Security” magazine