Recently, fraud cases caused by AI face-changing technology have attracted people's attention. With the development of large AI models, the application scope of "face-swapping" technology is becoming wider and wider. Technological breakthroughs have also made the facial expressions and action details of the characters after face-swapping more realistic. Once used by criminals, it may cause serious harm. . The reporter's investigation found that it is very easy to obtain AI face-changing software on the Internet. Where is the legal boundary of using AI face-changing software? How to do a good job in advance supervision?

01

Paid AI face-changing software appears on the Internet

Package prices range from 499 yuan to 2888 yuan

Currently, relying on breakthroughs in large models and algorithms, AI face-changing is being used in a wider range of applications, causing more concerns. Tutorials demonstrating the use and effects of face-changing software abound on the Internet. Different from the large number of free software for ordinary people to "try out", the reporter found that there are many high-priced paid AI face-changing software on the Internet. What is the difference between these paid software and free software?

After searching for "AI face-changing software", the reporter contacted a developer. He provided the reporter with a demonstration video of using his software to implement face-changing. The left side of the dual-window page is the original video before the face-changing, and the right side is On the side is a video after the face change. To the naked eye, the mouth shape, expression, and tiny movements of the "digital human" after the face change are almost flawless.

△Amateur face-changing effect shown by the seller (S is the original video, D is after the replacement)

On its sales website, the software developer also demonstrated how to splice the eyes and mouths of two well-known actresses together to generate a video of a new "digital human".

△The seller shows partial facial features replacement

This paid software offers different versions from basic (LV1) to advanced (LV4) to full set (LV5). For the entry-level package, 499 yuan provides software interface, video guidance, face-swapping models, etc. If you spend 2,888 yuan, you can get a full set of (LV5) video face-swapping models.

△A variety of face-changing software products of different levels are available on the Internet

02

"You can change any face you want"

AI large model can integrate multiple functions to create a “digital human”

Open the website,

There is no text at the bottom showing that it has been registered. The seller is also very cautious and only uses text to communicate throughout the process. He said that compared with the junior (LV1) to super (LV4) packages, the full (LV5) version has more iterations and stronger capabilities. It can also be understood that the face-changing effect is more realistic. At the same time, can not only achieve video Face changing also supports live face changing.

The reporter further asked, can the face be changed according to one's own needs? The software developer said: "You can change whatever face you want and use the materials for training."

△Screenshot of software developers chatting with reporters

A technician engaged in AI development said,

This means that there may be more risks that are difficult to control. "AI face-changing technology similar to deepfake is open source. Recently, large AI models have become popular, and its algorithm has been updated. It can integrate Vincent pictures, Vincent texts, Vincent audio, Vincent videos, and face-changing , combining such functions, can create a completely new 'digital human' or 'AI human', and the cost and difficulty of supervision will be several orders of magnitude higher."

△User feedback displayed by the seller

03

Expert: Effective labeling supervision system

Be able to clarify in advance the blurred boundaries caused by AI face-changing

Zhao Zhanzhan, a lawyer at Beijing Jiawei Law Firm and an IT law expert, said that from the perspective of post-attack, the criminal law has provisions on the crime of fraud; the civil code clearly states that citizens enjoy the right to portrait, name, reputation, etc. and other corresponding civil rights. In other words, if there is a related case of using other people's portraits for fraud, the criminals can be legally restrained and sanctioned afterwards. But with the rapid development of artificial intelligence technology, we should also explore how to achieve effective prevention in advance.

For example, the "Administrative Regulations on Deep Synthesis of Internet Information Services" requires that if the deep synthesis service may cause public confusion or misunderstanding, the deep synthesis service provider shall place the generated or edited information content in a reasonable position, Areas are clearly marked.

The "Measures for the Administration of Generative Artificial Intelligence Services (Draft for Comments)" requires that before using generative artificial intelligence products to provide services to the public, should report to the national cybersecurity and informatization department for a security assessment in accordance with relevant regulations; The legitimate interests of others must be respected; product providers should be responsible for the legality of pre-training data for generative artificial intelligence products, optimization of training data sources, etc.

Pan Helin, co-director of the Digital Economy and Financial Innovation Research Center of the International Joint Business School of Zhejiang University, believes that from now on, if the labeling requirements can be effectively implemented and an effective regulatory system can be formed, it can be achieved to a large extent. to clarify the blurred boundaries brought about by AI face-changing in advance.

Pan Helin: The current identification supervision system for AI face-changing has not yet been formed. Unlabeled content must be identified through technical means, and unlabeled individuals or individuals must be identified. The unit organization must have specific punishment measures, standards and systems.

Zhao Zhan believes that In the long run, supervision and punishment measures need to be continuously "iterated" with technological development in order to form a three-dimensional and effective governance plan.

Source: Voice of China WeChat official account

Reprinted from: People’s Daily Client

The above is the detailed content of AI face-changing software is sold publicly online, where are the legal boundaries?. For more information, please follow other related articles on the PHP Chinese website!