Click the topic to read previous chapters:
Chapter V Trade Secret Protection of AI
5.1 Typical Cases of AI Trade Secret Protection
The protection of trade secrets mainly refers to the judicial protection of trade secrets. The cases of judicial protection of trade secrets reflect the conditions and characteristics of the judicial protection of trade secrets to a certain extent, and is of great significance to enterprises’ management and protection of trade secrets.
Analysis of typical cases of AI trade secret protection is conductive to the quick grasp of its characteristics that are the same as and different from those of other technical fields.
5.1.1 Case Analysis
Case [(2017) Jing 73 Min Chu No. 2000] of dispute over trade secret infringement between Beijing Baidu Netcom Science & Technology Co., Ltd., Baidu Online Network Technology (Beijing) Co., Ltd. and Baidu (China) Co., Ltd. as the plaintiffs and Wang Jin as the defendant before Beijing Intellectual Property Court
On December 20, 2017, the plaintiffs Beijing Baidu Netcom Science & Technology Co., Ltd., Baidu Online Network Technology (Beijing) Co., Ltd. and Baidu (China) Co., Ltd. (collectively “Baidu”) sued Wang Jin to Beijing Intellectual Property Court for infringement of trade secrets.
Baidu's main causes of action is Wang Jin, former general manager of Baidu’s autonomous driving business department, who had already resigned at the time of the litigation, has carried out the following actions during his employment and resignation, which constituted an infringement upon Baidu’s trade secrets: 1. Violating the non-competition agreement and creating a company that directly competes with Baidu; 2. Failing to complete the procedures for handover at the resignation, to hand in computers, printers and other items storing Baidu’s trade secrets on the grounds of “loss”, and causing Baidu’s trade secrets to be infringed upon; 3. Registering companies in China and the United States before resigning, and poaching the core employees of Baidu’s autonomous driving business department.
The main claims of Baidu are as follows: Immediately stop the infringement upon Baidu’s trade secrets, including but not limited to stopping using such trade secrets to engage in autonomous driving-related businesses that compete with Baidu; Order the defendant to compensate Baidu for its economic losses and reasonable expenses of RMB 50 million, to make a public announcement to eliminate the impact, and to fully bear the legal costs.
After accepting this case, the court held a private hearing on this case on May 7, 2019. However, before the judgment was made, Baidu applied to the court for withdraw the lawsuit in writing, and this case was finally closed by way of withdrawing the lawsuit.
Criminal judgment made by a Shenzhen Court on the case of DJI’s source code leakage.
At the beginning of 2019, a Shenzhen court made the first-instance judgment on the case of DJI’s source code leakage. As the former employee of DJI volunteered to confess his guilty and showed repentance, the court sentenced him to six-month imprisonment for the crime of infringing upon trade secrets and a fine of RMB 200,000 after comprehensively considering the specific circumstances.
‘The cause of the incident was that when the employee was responsible for developing the code for the two modules of the agricultural drone management platform and the agricultural machinery spraying system in DJI, he uploaded the template code to the “public warehouse” on GitHub, causing the leakage of the source code. These code have been used in DJI’s agricultural drone products, and thus are trade secrets. Until 2017, security researcher Kevin Finisterr discovered this serious vulnerability in DJI. This vulnerability can enable attackers to obtain the private key of the SSL certificate and allow them to access sensitive customer information stored on the DJI server, thus making all the old keys of DJI useless, and such private data as user information and flight logs on the DJI server available for download. The leakage of the code caused an economic loss of RMB 1.164 million to DJI.
5.1.2 Characteristics of AI trade secret protection reflected in the case
Case I mentioned above is known as “the first driverless vehicle case in China”, on which the court did not make a formal judgment due to Baidu’s withdrawal of the lawsuit in the end, therefore, there is no authoritative information disclosed about the facts of this case. However, according to a statement later published by Allride AI Technology Co., Ltd., an affiliate of Wang Jin, this case was “withdrawn by Baidu in the end after twice trials and immediately before the pronouncement of the court” and “in this case, Baidu can neither prove the existence of trade secrets, nor prove Wang Jin’s infringement upon trade secrets”. This statement may imply that Baidu’s rights protection process is relatively difficult on the one hand; and the difficulty of this case, as an AI trade secret case, is exactly the same as that of a general trade secret case on the other hand, that is, the proof of ownership and infringement.
Case II mentioned above is a criminal case, in which the subject matter of trade secrets involved is the source code of software. Although AI field belongs to a new technical field, when seeking protection of trade secrets, the subject matter of existing trade secrets is protected first in practice. Therefore, the protection of AI, as a specific protection type, also has the common characteristics of the protection of the subject matter of trade secrets.
The subject matter of AI trade secret protection can be determined based on the laws currently in force and the specific characteristics of AI (see the foregoing content of this Report). Article 1 of the Provisions of the Supreme People’s Court on Several Issues Concerning the Application of Law in the Trial of Civil Cases Involving Infringements upon Trade Secrets stipulates that a people’s court may determine the information on structure, raw materials, components, formulas, materials, samples, styles, propagation materials of new plant varieties, processes, methods or their steps, algorithms, data, computer programs and their relevant documents, among others, relating to technology as technical information set forth in Paragraph 4, Article 9 of the Anti-Unfair Competition Law. The above-mentioned types of subject matter are common in know-how protection. Specifically, in AI field, technology-related algorithms, data, computer programs and related documents are the main subject matters of AI trade secret protection.
5.2 Conditions and characteristics of AI trade secret protection
5.2.1 Basic concepts of trade secret protection
Paragraph 4, Article 9 of the Anti-Unfair Competition Law of the People’s Republic of China provides that: Trade secret herein refers to commercial information such as technical information and business information that is not known to the public, has commercial value, and has been subject to appropriate confidentiality measures taken by the right holder.
This article officially defines trade secret in Chinese laws, and is generally accepted in theory and practice. According to the provisions of this Article, trade secret is essentially commercial information, with secrecy, value and confidentiality, that is, the “three features” in the constitutive elements of trade secrets generally referred to.
5.2.2 Conditions for Trade Secret Protection
The conditions for trade secret protection refer to the elements required for the constitution of an infringement and the obtainment of legal protection in case of any infringement upon trade secret. Article 14 of the Interpretation of the Supreme People’s Court on Several Issues Concerning the Application of Law in the Trial of Civil Cases Involving Unfair Competition stipulates that if a party alleges that others have infringed upon its trade secrets, such party shall be responsible for proving that the trade secrets it possess meet the statutory requirements, the information of the other party is identical or substantially identical to its trade secrets, and the other party has adopted improper means. Article 12 stipulates that any trade secret obtained through self-development or reverse engineering shall not be deemed to have infringed upon the trade secrets under Paragraphs (1) and (2), Article 10 of the Anti-Unfair Competition Law.
According to the above provisions, there are mainly four conditions for the protection of trade secrets, of which three are positive and one is negative. The three positive conditions are respectively constitute trade secret, access, and substantially identical, and one negative condition is not obtained through self-development or reverse engineering, which are commonly referred to as the constitutive elements of trade secret infringement: Constitute trade secret + access + substantially identical - legal source.
Among the above conditions for trade secret protection:
In terms of constituting trade secret, evidence that a trade secret meets statutory conditions includes the carrier, specific content, commercial value, and specific confidentiality measures taken for the trade secret, and others.
In terms of access, the court may consider the following factors related to the employees and former employees while determining whether such employees and former employees have channels or opportunities to obtain the right holder’s trade secrets: (I) Positions, responsibilities, and authorities; (II) work undertaken or the tasks assigned by company; (III) the specific circumstances of participating in production and business activities related to trade secrets; (IV) whether to keep, use, store, copy, control or otherwise access and obtain trade secrets and their carriers; (V) other factors that need to be considered.
In terms of substantially identical, if there is no substantial difference between the alleged infringing information and the trade secret, it can be determined that the alleged infringing information and the trade secret are substantially identical as set forth in Paragraph 2, Article 32 of the Anti-Unfair Competition Law. The following factors may be considered when the court determines whether they are substantially identical as mentioned in the preceding paragraph: (I) The degree of similarities and differences between the alleged infringing information and the trade secret; (II) whether those related people in the same fieldcan easily think of the difference between the alleged infringement information and the trade secret when the alleged infringement occurs; (III) whether the alleged infringing information and the trade secret are substantially different in the usage, use method, purpose, effect, etc.; (IV) information related to trade secrets in the public field; (V) other factors that need to be considered.
In terms of legal source, reverse engineering refers to the disassembly, surveying & mapping, and analysis of products obtained from public channels through technical means to obtain relevant technical information of the products. After the parties have learned of the trade secrets of others by improper means, they claim that the acquisition is legal on the grounds of reverse engineering, which does not belong to reverse engineering.
5.2.3 Characteristics of trade secret protection
Since AI field mainly involves the protection of technical information, compared with technical information, the protection characteristics of trade secrets are mainly reflected in the comparison with patent protection, which can be specifically divided into the following aspects:
First, the starting point for trade secret protection is lower than patent protection. The legal requirements for the protection degree of trade secrets are mainly reflected in the secrecy in the constitutive elements. The so-called secrecy is usually not known to the public. Generally speaking, information that is claimed as a trade secret can be deemed to meet the requirements of secrecy as long as it reaches a level that is not generally known to the public, without any need to meet higher requirements for inventive step.
Second, the protection of trade secrets does not require complicated authorization procedures. Trade secrets will be protected by law as long as they meet the constitutive requirements, without any need to apply for authorization to the competent authorities, and without determination and authorization by authorities on whether the trade secrets are constituted, which can be determined in specific cases by judicial authorities only when the judicial protection is sought for infringement upon trade secrets. Generally speaking, the unawareness of the public will naturally bring conditions for the right holders to exercise their rights, that is, a certain degree of exclusive right.
Third, there is no time limit for the protection of trade secrets. Trade secrets can be authorized without application, and the law does not specifically limit the duration of their rights. Accordingly, trade secrets can obtain continuous protection provided that they still meet the constitutive requirements stipulated by law.
Finally, trade secrets are more complicated in terms of management and rights protection. The aforementioned non-disclosure characteristic of trade secrets has not only brought advantages to the protection of trade secrets, but also brought more uncertainty to the management and rights protection of trade secrets. Due to the non-disclosure characteristic, trade secrets lack sufficient right appearance, and it is difficult for right holders to effectively delimit the boundaries of rights. On the one hand, it makes the implementation of confidentiality measures in management more difficult, and on the other hand, it increases difficulty in providing proof of trade secrets in rights protection.
It can be seen that whether to use trade secrets to protect technical information has advantages and disadvantages, and it depends on the company’s own situation and the characteristics of the information to be protected.
5.3 Suggestions for AI trade secret protection
5.3.1 General suggestions for AI trade secret protection
First, it is necessary to sort out trade secrets, mainly from the two aspects of secrecy and value.
Secrecy also refers to unknown to the public. Generally speaking, the information requested to be protected by the right holder shall be deemed to be unknown to the public as set forth in Paragraph 4, Article 9 of the Anti-Unfair Competition Law if it is not generally known and easily available to those skilled in the art when the alleged infringement occurs. Because it is negative information, secrecy also has negative conditions, that is, (1) the information belongs to common sense or industry practice in the field; (II) the information only involves the size, structure, material, simple combination of components, etc. of products, and can be obtained directly by those related people in the same filed by observing the marketed products; (III) the information has been publicly disclosed in public publications or other media; (IV) the information has been disclosed through seminars, exhibitions, etc.; (V) the information can be obtained by those related people in the same filed from other public channels. However, if the new information formed after sorting, improving, and processing the information known to the public conforms to the constitutive requirements for trade secrets, it shall be deemed as unknown to the public.
In terms of value. Generally speaking, the information requested to be protected by the right holder shall be deemed to have commercial value set forth in Paragraph 4, Article 9 of the Anti-Unfair Competition Law if it has actual or potential commercial value because it is not known to the public. If the phased results formed in the production and business activities meet the requirements, the results may be determined to have commercial value.
Second, it is necessary to manage trade secrets, mainly from the aspect of confidentiality measures.
In terms of confidentiality measures. In order to prevent the disclosure of trade secrets, the reasonable confidentiality measures taken by the right holder before the alleged infringement occurs shall be determined to be the corresponding confidentiality measures set forth in Paragraph 4, Article 9 of the Anti-Unfair Competition Law. When making determination, it shall be determined that whether the right holder has taken corresponding confidentiality measures according to such factors as the nature of the trade secret and its carrier, the commercial value of the trade secret, the degree of identification of the confidentiality measures, the degree of correspondence between the confidentiality measures and the trade secret, and the confidentiality intention of the right holder. If one of the following circumstances is sufficient to prevent the disclosure of trade secrets under normal circumstances, it shall be determined that the right holder has taken corresponding confidentiality measures: (I) signing a confidentiality agreement or stipulating confidentiality obligations in the contract; (II) requiring employees, former employees, suppliers, customers, visitors, etc. who can access and obtain trade secrets to keep confidential by means of articles of association, training, rules and regulations, written notification, etc.; (III) restricting visitors or conducting differentiated management on secret-related plants, workshops and other production and business places; (IV) differentiating and managing trade secrets and their carriers by means of marking, classifying, isolating, encrypting, sealing, and restricting the scope of persons who can have access to or obtain the trade secrets; (V) taking measures such as prohibiting or restricting the use, access, storage, copying, etc. of computer equipment, electronic equipment, network equipment, storage equipment, software, etc. that can have access to and obtain trade secrets; (VI) requiring the resigned employee to register, return, clear, destroy the trade secrets and their carriers that they have accessed or obtained, and continue to assume the obligation of confidentiality; (VII) taking other reasonable confidentiality measures.
5.3.2 Specific Suggestions for AI Trade Secret Protection
Article 1 of the Provisions of the Supreme People’s Court on Several Issues Concerning the Application of Law in the Trial of Civil Cases Involving Infringements upon Trade Secrets stipulates that a people’s court may determine the information on structure, raw materials, components, formulas, materials, samples, styles, propagation materials of new plant varieties, processes, methods or their steps, algorithms, data, computer programs and their relevant documents, among others, relating to technology as technical information set forth in Paragraph 4, Article 9 of the Anti-Unfair Competition Law. The people’s court may determine that information related to business activities, such as creativity, management, sales, finance, plans, samples, bidding and tendering materials, customer information, and data, constitutes the business information set forth in Paragraph 4, Article 9 of the Anti-Unfair Competition Law. The customer information referred to in the preceding paragraph includes the customer’s name, address, contact information as well as transaction habit, intention, content, and other information.
According to the said provisions and based on the characteristics of AI, the types of AI trade secret protection can be divided into know-how and business secrets. However, even for AI, the business secrets lack uniqueness, and have more commonality with other fields, so unnecessary details will not be given about the protection of business secrets. Protection suggestions will be discussed as follows mainly around several common technical scenarios of AI:
First of all, about the trade secret protection of AI data. Data is the foundation of AI development, and there are many types of AI data. According to whether the data is generated with the creative work of the right holder, the data can be divided into basic data and processed data. For basic data, more value is reflected after quantitative integration since it has not undergone any processing by the right holder, so such data can be protected as trade secrets as a whole. For processed data, both individual data and overall data can be protected as trade secrets since they may reflect certain protection values.
Second, about the trade secret protection of AI algorithms or programs. AI algorithms are valuable technical information. The algorithm or the program that implements the algorithm, that is, its related documents, can be protected as trade secrets, especially the source code that carries the algorithm, which can be protected as trade secrets without being made public.
Finally, about the trade secret protection of AI programs. There is little controversy about the protection of AI programs as trade secrets, and even in the AI field, the characteristics of the programs are basically the same as those in other fields from the perspective of trade secrets, so the AI programs protected as trade secrets are basically the same as other programs.