The latest target of China’s tech regulation blitz: algorithms

BEIJING — Chinese authorities are planning to restrict how companies use algorithms to sell products to consumers, a move analysts said likely runs counter to business interests and sets a precedent for other countries.

China’s largest tech companies from e-commerce giant Alibaba to TikTok-owner ByteDance have built their multibillion dollar businesses on algorithms that serve up content a customer is more likely to spend money or time on, based on previous viewing records.

The increasingly powerful cybersecurity regulator on Friday released sweeping draft rules for regulating use of these so-called recommendation algorithms. The proposal is open for comment until Sept. 26, with no specified implementation date so far.

The groundbreaking rules could set up a clash between China’s technology giants — which have been subject to increasing regulation over the past 10 months — and Beijing, which has sought to rein in their power.

And China’s algorithm rules will be closely watched by other countries and technology firms around the world for how it might affect business models and innovation, analysts said.

“Companies are going to have a lot to say about this because this has the potential to restructure business models,” Kendra Schaefer, Beijing-based partner at Trivium China consultancy, told CNBC.

The rules have also thrown up questions about how enforcement will happen and how intrusive regulators might have to be to actually get companies to comply with these rules.

Here are some of the key points in the draft rules:

• Companies must not set up algorithms that push users to become addicted or spend large amounts of money.
• Service providers need to notify users in a clear way about the algorithmic recommendation services they provide.
• Users need to have a way to switch off algorithmic recommendation services. Users should also have a way to choose, revise, or delete user tags used for the recommendation algorithm. 
• When algorithms are used to market goods or provide services to consumers, the company behind it must not use the algorithm to carry out “unreasonable” differentiation in terms of prices or trading conditions.
• Any violations of the rules could land companies with fines between 5,000 yuan and 30,000 yuan ($773 and $4,637).

These proposed rules come as the Chinese government has ramped up its regulation on homegrown technology giants in the last year, primarily in the name of cracking down on monopolistic practices and increasing data protection.

On Wednesday, a new data security law took effect. A personal data privacy law is set to take effect on Nov. 1.

Recommendation algorithms are formed of code that is fed specific information about users to help provide more tailored results. If you’re on an e-commerce site, some of items you see on the homepage are likely there because of your browsing or shopping habits.

But the algorithm’s code is not something that is made public and that could make enforcement difficult. At the very least, it could require regulators to inspect companies’ code behind the algorithms.

“You can’t carry out algorithmic regulation without looking at the code,” Trivium China’s Schaefer said.

Authorities are to carry out algorithm “security assessments” and inspection of the recommendation services, according to the draft rules. Companies must cooperate and provide any necessary technical or data support.

That would give regulators in China enormous power.

But it also throws up some challenges.

“First of all you need the technical capacity to do this. … You also need the bureaucratic process to do it. All that has to be sorted and it has not been yet,” Schaefer said.

This intrusiveness could set up a clash between China’s technology giants and regulators.

“I’m sure there are issues with privacy rights with companies … that [the code] is proprietary information,” Schaefer added.

None of the Chinese tech companies contacted by CNBC had immediate comment on the draft rules, with two indicating it’s too early in the process to assess them. The cybersecurity regulator did not immediately respond to a CNBC request for comment on the extent of implementation or impact on innovation.

Many of China’s technology giants aren’t making money off of their algorithms directly. Instead, they’re used to direct consumers to products. For example, you may be watching videos on an app and then get recommended similar content. A company would monetize that via advertising or even getting you to buy things.

The latest rules could have the potential to force companies to change their business models, but it’s unclear as to what extent.

“The jury is still out on the implications for operations and profits,” said Ziyang Fan, head of digital trade at the World Economic Forum.

“It depends on a number of factors, such as the level of enforcement, and market reactions — how many users would choose to ‘turn off’ [the] recommendation algorithm if that’ll lead to a suboptimal user experience, such as getting cat videos pushes when you are a dog person?” he said in an email.

“If we see a significant drop in indicators such as DAUs [daily active users] and retention rates, then the implications for profits could also be significant,” he said, noting that social media companies may see the impact more, while online shopping and ride-hailing “probably less so.”

As the intersection between tech and daily life grows, countries and regions around the world are increasingly looking at ways to regulate technologies and the companies that sell them.

That’s resulted in different approaches, so far. In the area of algorithms, China is specifically focused on the technology’s recommendation feature, while the U.S. and European Union are discussing broader laws around artificial intelligence.

Earlier this year, the European Union issued a draft law called the Artificial Intelligence Act with the purpose of facilitating “the development of a single market for lawful, safe and trustworthy AI applications” and pushing innovation in the space.

The law has “specific requirements that aim to minimise the risk of algorithmic discrimination.”

But there are a number of differences with China’s algorithm rules.

WEF’s Fan said the EU follows a “risk-based approach” while China’s rules “do not differentiate risk levels and apply to all use of algorithm recommendation technology.” That can cover a broad range of industries from food delivery to education.

And China’s rules “target algorithms directly at the user and product level,” such as the ability for users to switch off the algorithm, as stated in the proposed rules, Fan added.

Once enacted, China’s law on algorithms will be closely watched around the world as authorities try to figure out how to regulate technology in the future.

“This is going to set a global example,” Schaefer said. “Tech companies overseas are going to see how Chinese tech companies do or do not profit given these restrictions on algorithms. If they change business models, if they can succeed despite regulation on algorithmic process, there is very little excuse for … foreign governments not to do the same.”

“If they fail and they are not as profitable and shareholders are disappointed, then that is bad, too,” she said. “That bolsters the argument you can’t implement algorithmic regulation without detrimental effects to innovation.”