From recommending personalised contents on social media platforms to designing work schedules for motorcycle couriers, algorithms are critical to many technology companies’ operations. However these “black boxes” of Big Tech are under increased regulatory scrutiny in China with the advent of a new algorithm-specific regulation for the world’s second largest economy – the Internet Information Service Algorithm Recommendation Management Regulation (Regulation).
At the intersection between data and anti-trust laws and ethics principles, although the technical rules comprising the Regulation will prove far-reaching in shaping business models driving China’s internet services, question how long they will remain unique to the Asia superpower. Even if your tech operations are not based in China, the commentary below may give you a head start on what is coming down the (digital) track.
Building on China’s cyberspace framework
With China’s three pillar laws regulating its massive cyberspace now in effect from late last year – (1) the Cybersecurity Law, (2) the Data Security Law, and the (3) Personal Information Protection Law – the market continues to wait for more detailed regulations that refine and implement the broad provisions of this framework. As for other markets in which we advise, China has seen growing concerns among its netizens around issues like algorithm discrimination, a lack of transparency from Big Tech, and increasingly acute labour grievances caused by the use of algorithms to schedule work.
In early January, the Cyberspace Administration of China (CAC) issued the Regulation to provide some of that detail. The Regulation came into effect on 1 March and the CAC launched a campaign from April to the end of 2022 to ensure early implementation of the Regulation.
Key obligations to focus on
- Security responsibility and regular review. The Regulation makes operators deploying algorithms with recommendation functionality (Operators) the first party responsible for the operation and results from using those algorithms. Operators must establish systems and technical measures to review algorithms from both a technology and ethics perspective. Considering the proximity of the results generated by algorithms and the input data, training models and application scenarios, the Regulation also specifies that Operators must carry out daily monitoring on algorithms’ operations and outcomes to predict potential safety issues and ethical problems in advance. Where illegal information is discovered, Operators must prevent it from spreading, in line with obligations given a statutory footing in Cybersecurity Law. The “S” in ESG is a clear foundation to the Regulation.
- Right to know and right to choose. Under the Regulation, Operators should conspicuously inform users of the use of algorithmic recommendations and publicise the basic mechanisms behind them. The Regulation also provides for the protection of user rights including the provision of options/information that are not based on their personal characteristics, convenient opt-outs to turn off algorithms, and the ability to select or delete user tags fed into these algorithms. These obligations should force greater transparency by Operators in their deployment of algorithms and ultimately allow users to gain a clearer understanding of how their data is used. Tech and design teams of Operators will need to collaborate to reconfigure user interfaces with the required functionality, adding to compliance time and capital costs. Indeed, by mid-March, many of the most popular apps in the Chinese market had adopted a button for users to turn off algorithm-based recommendation services.
- Algorithm’s administration and filing requirement. To measure the relative importance of algorithms to the market and strengthen and refine its supervisory practices accordingly, the CAC will establish an administrative system that classifies algorithms by properties like their content type, number of users, and their impact on public opinion. Specifically, the Regulation requires algorithms which boast recommendation functions which could “shape public opinion” or “mobilise society” to file with the CAC a submission pack that includes a self-assessment report. So far it appears that no companies have completed the filing for their algorithms.
Tackling pressing societal and antitrust concerns
The Regulation requires Operators to design their algorithms to adhere to “orientation of mainstream values” and promote “positive content”, societal concepts inherited from the Data Security Law. It also responds to some problems cited as being prevalent in today’s Chinese society, such as:
- teenage addiction to the internet, by requiring Operators to consider teenagers’ physical and mental health and not use algorithms to induce minors to over-indulge in online activities; and
- the elderly’s difficulties in using internet services, by mandating that appropriate measures must be adopted by Operators to facilitate seniors’ use of algorithm recommendation services.
In addition to these societal requirements, the Regulation also pays considerable attention to antitrust principles, which have also been vigorously enforced within China’s digital economy for 12-18 months. The Regulation makes it clear that using algorithms to engage in anti-competitive practices or price discrimination is prohibited.
What do businesses need to do
While regulators may take time to enforce the Regulation – since violations may not be overtly obvious without systematic use of websites and mobile applications – considering the costs and resources needed to update policies and systems in compliance with the Regulation’s broad-ranging requirements (as well as the ongoing need to monitor how it will be implemented in practice), we foresee that many Operators will need time to fully adapt to the new rules.
However, given the growing trend towards users exercising their virtual rights and filing complaints against non-complaint digital service providers, as well as the fact that the Regulation provides clear guidance around the design and operation of algorithms, Operators should seek to develop a feasible compliance solution in line with its requirements – and those of other recently published tech regulations – as soon as possible. Avoiding a consumer backlash, and potential regulatory sanction, is a reality of operating in China’s increasing scrutinised digital landscape.
(Thanks to our researcher Xinyi Hua for her support in preparing this article.)