Huya and Douyu: how abusive complaints undermine the safe harbour, and what a platform can do

Introduction

Douyu complains again and again, Huya answers each time, and Apple is caught in the middle. How should a platform build an effective mechanism for handling complaints?

The internet has been following one matter closely. Huya, unable to bear Douyu's repeated complaints to the Apple App Store, brought a suit for unfair competition in the people's court of Nansha district, Guangzhou, and obtained an order for conduct preservation requiring Douyu to stop at once complaining to the App Store and to Apple about Huya. It is the first injunction known in China against complaints made to the Apple App Store.

Background

From 28 August 2018 Douyu began complaining to Apple about Huya. The substance of the complaints was that Douyu had concluded agency contracts with the streamers Zhang Fuhai, Wu Xinyuan and Yu Xing under which the copyright in audio-visual works made by the streamers during the term of the contracts belonged to Douyu; that Huya, without Douyu's authorisation, was disseminating those works and infringing Douyu's copyright; and that Apple should remove the infringing Huya app from its store.

According to the ruling ((2019) Yue 0115 Min Chu No. 1339), between 28 August 2018 and 15 February 2019 Douyu complained to Apple about Huya on the same matter 23 times. Huya supplied proof of its contracts with Zhang Fuhai, Wu Xinyuan and Yu Xing, together with statements that it had no cooperation with Douyu or that such cooperation had ended, but Douyu went on complaining to Apple.

What role did the App Store play in all this?

When a complainant files a complaint, the App Store, as a distribution channel for applications, sends the complained-of party an email notice containing the complainant's identity, the grounds of complaint, the suggested handling and the consequences. Throughout, however, Apple makes essentially no substantive judgment of its own.

The complained-of party may then approach the complainant directly, or reply by email setting out its position, and Apple forwards that reply to the complainant. The complainant may then complain again, and so the cycle repeats. The process is shown in the diagram below.

To many observers Apple's approach displays the Chinese culture of the middle way to the full, and has very much the air of standing on the bank watching the tiger and the fish fight.

Apple, as the platform, is caught in the middle: is the safe harbour a harbour or a cage?

The safe harbour principle

Anyone who follows the internet or the law knows the "safe harbour" principle, also called the "notice and takedown" rule. Where a user of network services commits a tort, the injured party may notify the service provider to take necessary measures such as deleting, blocking or disconnecting links. If the provider fails to act promptly and the injured party's loss grows, it is jointly liable with the tortfeasor for the increase in the loss; if it acts promptly after receiving notice, it bears no liability and sails safely into harbour. The reason is that the rise of the platform economy makes it objectively impossible to review the vast quantity of content involved.

The safe harbour in brief

The safe harbour principle is not, however, a panacea. It is itself no more than a compromise between fairness and efficiency, and first appeared in the Digital Millennium Copyright Act of the United States in 1998.

The E-commerce Law, which entered into force in China only recently, further refines the principle. It applies at present only to e-commerce, but some of its rules are worth borrowing: where the platform operator has forwarded a statement that no infringement exists and within fifteen days has received no notice that the right holder has complained to the authorities or brought a suit, it shall terminate the measures against the complained-of party without delay; double compensation for a mistaken complaint; and the measures available to a platform include, besides deletion and blocking, termination of transactions and services, such as deducting points, closing the shop or barring the complained-of party from promotional events. Introducing the safe harbour is not, of course, the point of this article.

Returning to this case: Apple did not remove Huya at once, and Huya supplied counter-evidence showing that it had not infringed, yet Douyu insisted on complaining again and again while Apple continued to play the firm neutral.

A platform that makes no judgment at all and leaves the two sides to argue it out does minimise its own risk, but in an age that prizes service and efficiency it must also damage its competitiveness.

Apple, in truth, has difficulties of its own. In its dispute with Mai Jia over copyright ownership and infringement (case number (2013) Gao Min Zhong Zi No. 2619), the courts at both levels in Beijing came close to denying the App Store its status as a provider of information storage space. That means the App Store can hardly invoke the safe harbour, and will become the target of right holders seeking remedies, which tests the capacity of Apple's legal department and its ability to solve problems.

Among the ways of handling complaints, the purely neutral mechanism and the remove-on-notice mechanism may fairly be ranked as the two worst.

How, then, is a platform to balance effectiveness against risk?

Several dimensions may be considered.

1. A platform is not a professional tribunal of disputes, and should not be held to a highly specialist standard.

2. The safe harbour principle is not a principle of attribution but a condition of exemption, and not a sufficient and necessary one. It is not that liability follows where notice is given and nothing is removed, for it is not a condition of attribution; nor that liability is avoided where something is removed after notice, for the red flag principle applies, that is, the platform knew or ought to have known.

Matters of this kind are in essence tort disputes, and whether a platform bears tort liability should be governed by fault. A court should not conclude that a platform is at fault simply because nothing was removed after notice; it should assess whether the measures the platform took discharged its duty of care. That question returns to point 1: a platform is not a professional body for resolving disputes, and it should not be judged too harshly. Whether the duty of care was discharged must be analysed case by case in the proceedings.

3. Make good use of Article 36, paragraph 2, of the Tort Liability Law.

"Where a user of network services commits a tort, the injured party has the right to notify the service provider to take the necessary measures such as deleting, blocking or disconnecting links. Where the service provider fails to take the necessary measures promptly after receiving notice, it is jointly liable with that user for the increase in the loss."

The words "necessary measures" open up many possibilities for a platform, and how it uses them shows its wisdom. The law should also allow a platform to use a range of instruments to deal with such matters, so as to meet abusive complaints and other situations, such as infringers who exploit the interval between notice and takedown to infringe content whose value is short-lived.

The cost of review and handling also deserves attention. If a platform does no more than Apple did here, acting as a relay for emails, a machine could do the job at negligible cost. To be effective while keeping risk under control, however, necessarily consumes a great deal of the platform's legal resources: a small volume of complaints can be handled, a large volume cannot. That is why many platforms adopt a wholly passive approach.

Looked at from the point of view of cost, however, judging effectively and promptly, bringing a complaint to an end, and directing the parties to the courts is another way of saving money.

On that analysis it is worth exploring how a platform should build its own complaint-handling mechanism.

Suppose you run a video platform and face a thousand complaints a day. How should you deal with that volume while preserving the efficiency of your service?

(1) Should the guiding principle be the user experience or risk control?

From the point of view of commercial development the user experience should come first, and risk should then be reduced to a minimum on that basis.

(2) How should the assessment process be built?

1. Artificial intelligence should filter out some complaints first, and this is probably the first thing to consider. Set up an effective channel for filing complaints, for example: a complainant must provide complete proof of title, identity and contact details, and without that information no effective complaint can be filed.

↓

2. A first-tier "sorting" review team takes the first round and decides quickly whether to remove the content.

↓

3. Content marked for removal goes to a second-tier "user experience" team, which makes the second decision on removal.

↓

4. Content that both rounds decide to keep goes to the legal team, and if the legal team considers that it should be removed the matter returns to the user experience team.

On that basis a complaint assessment mechanism might be built as follows.

The assessment process

The flow chart shows that the power to remove rests in the end with the user experience team alone, a result driven by commercial interest.

If risk control is thought more important, the power to remove may be placed in the hands of the legal team. An enterprise may adopt different strategies at different stages: in preparation for an IPO, for instance, the latter may be preferred.

Where the power to remove lies with the user experience team, the demands on that team are higher, since it needs a feel for commercial operations. In this model the legal team has no power to decide removal, but a decision of its own that content should be removed becomes one of the reference points by which the user experience team weighs the matter again.

The model can be refined further. What factors, for instance, should the user experience team weigh?

(3) Setting standardised assessment factors

The factors can be listed item by item with weights assigned in advance, the content scored accordingly, and removal decided where the total falls below a figure, 60 for example.

A simple example follows.

The table above is illustrative only; the factors and their weights can be set from experience in individual cases, and the same method can be applied to the legal team.

A process-driven and standardised assessment improves both efficiency and consistency.

The foregoing is only a brief analysis of a platform's complaint-handling mechanism. Such a mechanism should be complex and involve several departments working together, and it must be refined in practice and in litigation until it matches the enterprise's own strategy.