The new regulations of "deep synthesis" and the key technologies of the "metacosm" usher in supervision

There is no doubt that artificial intelligence (AI), big data, blockchain and other technologies have profoundly changed our way of life in just a dozen years, in this increasingly virtual and real world, new things are constantly being created, and gradually form a larger and more complex network of relationships with the things of the old generation.

Deep Synthesis (formerly known as Deepfake, or "deep forgery") has attracted much attention in recent years as a new application and practice of AI technology.

On the one hand, the use of deep synthesis technology to automatically generate text, speech, images, videos and other digital content has become a key technology in the development of the current metaverse; on the other hand, in recent years, the rapid development of science and technology, the threshold of deep synthesis technology has been rapidly reduced, resulting in frequent cases of abuse of technology, causing various risks.

Therefore, it is urgent to regulate the application of deep synthesis technology in the form of legislation, especially to make specific provisions on the use, labeling, scope of use and penalties for abuse of technology of deep synthesis content, and at the same time clarify the responsible subjects, in order to truly achieve "technology for good".

On January 28, 2022, the Cyberspace Administration of China (CAC) recently published the Provisions on the Administration of Deep Synthesis of Internet Information Services (Draft for Solicitation of Comments) (hereinafter referred to as the "Deep Synthesis Consultation Draft"), which makes a series of clearer provisions and guidelines for deep synthesis technology as the cornerstone of the metacosmum. This article interprets this document to help readers understand how deep synthesis technology can be used to comply.

First, deep synthesis technology and its practical application in the metaverse

Technically, "deep synthesis" technology, which is essentially implemented using deep learning algorithms, mainly uses two technologies: autoencoders and generative adversarial networks (GANs).

One is used to generate new data, one is used to identify the generated data, and after countless "cooperation" between the two, a very "realistic" synthetic data is finally generated. The current familiar AI face swap is created by using deep synthesis.

According to Article 2 of the Draft for Comments on Deep Synthesis, deep synthesis technology refers to:

Technology that uses generative synthesis algorithms represented by deep learning and virtual reality to produce text, images, audio, video, virtual scenes and other information, including but not limited to:

(1) Techniques for generating or editing text content, such as chapter generation, text style conversion, and Q&A dialogue;

(2) Text-to-speech, speech conversion, speech attribute editing and other technologies for generating or editing speech content;

(3) Music generation, scene sound editing and other technologies for generating or editing non-speech content;

(4) Face generation, face replacement, character attribute editing, face manipulation, posture control, and other technologies for generating or editing biometric features such as faces in images and video content;

(5) Techniques for editing non-biometric features in images and video content, such as image enhancement and image restoration;

(6) Technologies for generating or editing virtual scenes, such as three-dimensional reconstruction.

According to this definition, AI voice, NFT generation art, virtual concerts, holographic portrait projection, virtual people digital people, AR shopping and other important components of the metaverse belong to the specific application of deep synthesis technology. As long as the application of deep synthesis technology to provide Internet information services within the territory of the People's Republic of China, or the provision of technical support for deep synthesis services, are within the scope of the "Deep Synthesis Consultation Draft".

With the gradual maturity of "deep synthesis" technology and entering the stage of commercial application, its huge economic value has gradually been revealed to the world. At present, "deep synthesis" has been actually promoted and applied in scenes in music, film and television, comprehensive entertainment, social networking and other industries.

For example, in the production of blockbuster movies, its application is the most mature, not only can improve the efficiency and quality of audio and video production, significantly reduce the cost of film production, but also give artistic creators more imagination space. Even deep synthesis technology can enable some deceased actors to be "resurrected" in the form of virtual digital humans in film and television dramas.

In terms of comprehensive entertainment, in-depth synthesis is also very good, and face effects and face-changing short videos commonly used by an Internet manufacturer have been commonly seen in our lives. Virtual idols, virtual anchors, and even those who already have the ability to provide consumers with entertainment, shopping, wizards and other services. In addition, in the fields of education, art, medical treatment and scientific research, deep synthesis technology can bring new applications to these industries.

Second, the risks of deep synthesis technology

Since the advent of Deepfake (Deep Forgery) in 2017, the potential of AI technology in counterfeiting has been amazing to the world. Since then, the rapid development of GAN algorithms can not only achieve AI face change, but also automatically generate various digital content such as text, artificial speech, and images.

Deep synthesis technology, while facilitating our lives, also brings a series of risks. For example, a large number of pornographic videos produced by AI face swaps are flooding the Internet, which not only seriously infringes on the personal interests of others' portrait rights, privacy and reputation rights, personal information security, etc., but also causes serious harm to social public order and good customs. There are even criminals who use AI face-changing, synthetic voice and other technologies to carry out criminal activities such as fraud.

In 2021, the Cyber Security Corps of the Shandong Provincial Public Security Department announced several real cases of fraud using deep synthesis technology.

In December 2020, a company's finance received a call from a leader requesting that it immediately transfer 20,000 yuan to the supplier and send the transfer information to the other party's mailbox in the form of an email, on the grounds of avoiding late fees. Because the boss's accent on the phone was so realistic, the company's finance completed the transfer within an hour without any doubt.

What's more, in the second case disclosed by the Shandong Cyber Security Corps, criminals used social media videos to defraud.

First of all, by stealing other people's social media accounts, obtaining other people's facial information, and asking relatives and friends to transfer money to their designated accounts by making video calls with the relatives of the original owners of the accounts, they successfully gained the trust of the victims and defrauded a large amount of money.

The use of deep synthesis technology to carry out illegal and criminal activities is the appearance of its risk, and the real danger of deep synthesis technology is its serious damage to social trust relations. The existence of trust relationships is extremely important for any individual and community, and even some scholars believe that trust is one of the foundations of human society since ancient times, and the importance of trust relationships between people cannot be overestimated.

Teacher Zhang Kangzhi believes that in the agricultural social stage of human history, the trust relationship between people is expressed in the form of traditional customs, while the trust relationship between people in modern industrial society belongs to the contractual trust relationship. Although the form of social trust will change with the times, in any case, the maintenance of society must be based on social trust.

Deep synthesis technology has the potential to undermine social trust, as the saying goes, seeing is believing, but when seeing and hearing are no longer true, lies and deception are rampant, social trust will disappear.

In his Critique of Practical Reason, Kant argued for the horror of a "trustless society." According to Kant's theory, if lies are the prevailing laws of society, then everyone can lie. In this case, as a person with normal reason, we will no longer believe the liar, and we will no longer believe everyone in this world. Lies that are not believed no longer have meaning, which forms a paradox.

It is precisely from this that Kant deduces that "not to lie" and social trust are the normal forms of social existence and development, because only in social trust can liars profit from lies.

Today, information networks have penetrated into all corners of our lives, and the unrestrained abuse of deep synthesis technology will inevitably bring us a society with chaotic value, loss of trust and disintegration. Therefore, it is urgent and necessary to regulate it reasonably in the form of law.

Third, the key points of compliance with deep synthesis technology

As the first regulation in mainland China to specifically regulate deep synthesis technology, the Draft for Comment on Deep Synthesis plays a vital role in the compliance development of technology companies such as Meta-Universe Enterprises.

(1) Responsible subjects and specific requirements

According to Article 2 of the Draft for Comments on Deep Synthesis, there are two types of subjects responsible for deep synthesis technology:

(1) Entities that apply deep synthesis technology to provide Internet information services within the territory of the People's Republic of China;

(2) The main body that provides technical support for deep synthesis services.

For these two types of entities, according to Article 7 of the Draft for Comments on Deep Synthesis, they need to:

Implement information security entity responsibilities, establish and complete management systems such as algorithm mechanism mechanism review, user registration, information content management, data security and personal information protection, protection of minors, and employee education and training, and have safe and controllable technical safeguard measures that are compatible with the development of new technologies and new applications.

In addition, Article 16 of the Draft for Comments on Deep Synthesis specifically stipulates the responsibilities of Internet application store service providers:

The deep synthesis application provided by the deep synthesis service provider shall perform safety management responsibilities, and verify the safety assessment, filing, and other circumstances of the deep synthesis application in accordance with the law; where relevant national provisions are violated, disposal measures such as not being on the shelves, suspending the shelves, or removing the shelves shall be promptly adopted.

(2) Procedural compliance

1. ICP Filing

According to Article 19 of the Draft for Comments on Deep Synthesis, deep synthesis service providers shall follow the relevant provisions of the Provisions on the Administration of Internet Information Service Algorithm Recommendation (hereinafter referred to as the "Algorithm Provisions").

Complete the filing formalities within 10 working days from the date of service provision. Deep synthesis service providers that have completed filing shall indicate their filing number and provide links to publicity information in a conspicuous position on the websites, applications, and so forth that they provide services to the outside world.

2. Security Assessment

According to Articles 11 and 20 of the Draft for Comments on Deep Synthesis, different types of safety assessments are required depending on the application of deep synthesis technology: (1) conduct safety assessments on their own; (2) conduct safety assessments in accordance with relevant national regulations.

For models, templates, and other tools that have the ability to edit non-biometric information such as faces and voices, or special objects or scenes that may involve national security or social public interests, they shall carry out security assessments on their own. For deep synthesis service providers to develop new products, new applications, or new functions with public opinion attributes or social mobilization capabilities, they need to conduct security assessments in accordance with relevant national regulations.

(3) Substantive requirements

Articles 8 to 14 of the Draft for Comments on Deep Synthesis set forth the requirements of the subjects responsible for deep synthesis technology, mainly including: information security, authentication of real identity information, technical review and assessment, data security assurance, content management, content identification and traceability, and content identification.

Among them, special attention needs to be paid to the content identification part, according to Article 14 of the "Deep Synthesis Draft for Comments":

Where deep synthesis service providers provide deep synthesis services, they shall use conspicuous methods to identify the content of deep synthesis information, and effectively prompt the public of the composition of the information content.

For violations of the "Deep Synthesis Draft for Comments", in terms of administrative responsibility, they will face the consequences of warnings, orders to make corrections within a time limit, suspend relevant business, suspend information updates, and impose a fine of between 10,000 and 100,000 yuan. In civil terms, if damage is caused to others due to violations of relevant regulations, civil liability shall be borne in accordance with law. In criminal matters, criminal responsibility is pursued in accordance with law for acts constituting a crime.

Fourth, write at the end

Deep synthesis technology is a typical scientific and technological "double-edged sword", only by delineating an appropriate framework for its use, can science and technology revitalize the country and benefit the whole people in the true sense.

Of course, no technology has emerged after human beings have been fully prepared, such as deep synthesis, big data, and blockchain technology. Today, when we are about to fully usher in the digital world, we must neither choke on food, restrict the development of deep synthesis with too strict legislation, nor allow it to be too loose. How to legislate more appropriately and reasonably is the responsibility of all our legal personnel to the times.