Faced with massive amounts of information, algorithms recommend how to keep boundaries

　　The widespread application of artificial intelligence and big data has brought automated algorithm technology to the Internet platform.

On the one hand, Internet platforms use algorithm technology to extract features and identify network content, and provide Internet users with literature, music, videos, etc. through tagging; Information such as reading habits, consumption preferences and other information are captured and analyzed, and relevant content is accurately recommended to network users.

The application of algorithms has injected new kinetic energy into economic and social development. At the same time, the unreasonable application of algorithms such as algorithm discrimination, "big data killing", and inducing addiction has also led to a series of problems.

　　A few days ago, the State Internet Information Office and other four departments jointly promulgated the "Internet Information Service Algorithm Recommendation Management Regulations".

From the perspective of public interests and consumer protection, the regulations make it clear that the algorithm recommended to users should not contain illegal information, which will help promote the healthy development of algorithm recommendation services and improve the level of supervision capabilities.

However, if the content recommended by the algorithm contains information involving the rights of a third party, how the rights of the third party are guaranteed is not covered in the regulations.

In order to maintain the Internet ecology and balance the relationship between Internet platforms, rights holders, and users, when establishing the obligations of Internet platform algorithm recommendation, we should reasonably grasp the neutrality of technology and the flexible interpretation of legal rules.

　　Popularize and popularize filtering technology on Internet platforms.

When the Internet platform automatically recommends content, it generally provides users with keywords, etc., according to the user's reading and appreciation habits.

The stronger the information capture and information analysis capabilities of the backend of the platform enterprise, the more similar the content obtained by users to what they have read and appreciated before.

If these contents are music and videos copyrighted by others, Internet companies may be liable for indirect infringement.

Some Internet companies will use technology neutrality as an excuse to claim that they only automatically identify through keywords, etc., and cannot see the content presented to users.

In order to avoid illegal and infringement disputes, Internet platforms should be encouraged to adopt algorithm filtering technology and establish and improve feature databases for identifying infringement and bad information.

Similar to algorithm recommendation, algorithm filtering can effectively filter out those suspected infringing texts or videos through the identification function, greatly reducing the risk of infringement.

In addition, if Internet users use the Internet platform to automatically recommend infringing articles to other users, in order to avoid their own liability risks, Internet platforms can also use algorithm filtering technology, supplemented by manual intervention mechanisms, to warn Internet users, or even take other penalties for it.

　　科学解释互联网领域的通知规则。从行为上看，互联网平台向用户提供的，要么是内容，要么是服务。如果互联网平台向用户提供的是文本、视频等内容，而且这些内容如果侵权，互联网平台就构成了直接侵权行为。如果互联网平台向用户提供自动存储空间、搜索、链接等技术服务，仅在主观上存在过错的前提下，承担间接侵权责任。互联网的本质是连接。为保证互联网行业的健康发展，民法典和其他民事法律规定了互联网企业的免责条件。如果权利人向互联网平台发出了符合条件的通知，告知互联网平台上存有侵害其权利的作品，互联网平台及时删除了该作品，及时断开链接或者采取了其他必要措施，互联网企业就不承担法律责任，这就是互联网领域所谓的“通知-删除”规则。随着算法技术的广泛应用，越来越多的第三方机构利用算法代表权利人向互联网平台发出侵权通知。例如，创立于2011年的某科技公司，截至2020年11月，已监测发现4700多万条侵权链接。为避免侵权风险，提供服务的互联网平台被迫采用算法技术对涉嫌侵权作品进行比对并删除，由此，互联网环境下的“通知-删除”规则变成了“算法通知-算法删除”规则。有学者指出，在这种模式下，由于机器人之间的对话取代了人与人之间的对话，只能将互联网平台的算法系统对“算法通知”的接收，“视为”平台的“知道”。

　　Reasonably establish the duty of care of the Internet platform.

Duty of care is the premise for Internet platforms to take responsibility.

When Internet platforms provide services, because they are faced with massive amounts of information, the law does not generally require them to undertake prior review obligations.

That is to say, when the Internet platform provides services, it does not review the legality of uploaded works in advance.

However, if the infringement of Internet users is very obvious, such as recommending popular film and television works by setting a list, and the user's infringement is as clear as a red flag, reaching the "red flag standard", the Internet platform has violated due attention. obligation, constitutes indirect infringement.

It should be noted that the act of recommending allegedly infringing works by an Internet platform using an algorithm cannot be classified as a traditional act of providing works because the content of the work cannot be seen in the background; Difference is an act in between.

Therefore, it is too harsh to make it take direct responsibility.

However, if the "notice-deletion" rule is generally applied, it is inconsistent with its initiative to recommend suspected works.

Article 12 of the Provisions of the Supreme People's Court on Several Issues concerning the Application of Law in the Trial of Civil Disputes Infringing the Right of Information Network Dissemination stipulates that if the Internet platform voluntarily edits the theme and content of the popular film and television works and provides services, it is considered that It violates the duty of care.

Although the Internet platform will edit and process keywords, etc. before the algorithm recommends the work, this is only machine editing, and it cannot be generally regarded as a violation of the duty of care.

In current practice, some rights holders protect works by adding video fingerprinting technology to some popular film and television works, so as to track the dissemination of infringing works.

If the right holder requires the video website that automatically recommends the work to adopt the fingerprint recognition technology, the video website may consider it has not fulfilled the necessary duty of care if it does not adopt the technology.

Of course, some countries and regions currently have stricter regulations on algorithmic recommendation behavior, requiring Internet platforms to confirm the copyright of videos shared online, which may inhibit the application of algorithmic recommendation.

　　(Author: Li Yufeng, Dean of Intellectual Property Research Institute of Southwest University of Political Science and Law, Chief Expert of Intellectual Property Governance Innovation Research Team)