A Six-Year Retrospective on the Defence Strategy in a Live Quiz Assistant Case (2020-2026)
Authors: Rong Chao and Gu Zihao, Attorneys at Law, Boss & Young Attorneys-at-Law, Shanghai.
Abstract
Between 2020 and 2022, the authors represented a search-engine enterprise in a dispute brought by an internet company concerning whether a “live quiz assistant” constituted unfair competition. The case was an early Chinese dispute addressing the assessment of an AI-assisted tool intervening in a third-party online activity, at a time when generative AI had not yet reshaped global legal debate and AI law was still in its infancy. The defence advanced three sets of core propositions: the nature of AI, namely that AI-assisted answering is functionally homologous with human visual reading, semantic understanding and search; the protection of maverick competitors, namely that disruptive AI tools should not be suppressed by conservative invocations of business autonomy; and a non-zero-sum market test, namely that the tool did not destroy the claimant’s commercial model but may have generated traffic dividends. Looking back from 2026, many of those arguments have become part of the basic vocabulary of AI law.
Keywords: artificial intelligence; Anti-Unfair Competition Law; maverick competitor; Schumpeterian competition; Turing test; non-zero-sum game; competitive neutrality; creative destruction.
I. Introduction: Why Revisit a Case from Six Years Ago?
In May 2026, when we reopened the cover pages of the defence submissions and civil appeal filed six years earlier, the true significance of that litigation became clearer than it had been at the time. From 2020 to 2022, we represented a search enterprise in an unfair-competition claim brought by an internet technology company. The focus of the dispute was an AI-assisted tool, which we described as a live quiz assistant. Through OCR image recognition, natural-language semantic analysis and big-data search, it provided users of the defendant’s search product with reference answers for a live quiz activity operated by the claimant.
Six years later, AI has moved from a peripheral assistance tool to a general-purpose intelligence infrastructure embedded in social life. Generative AI, agents and human-machine collaboration are no longer purely technical topics. They are legal objects that legislators, courts and scholars must address directly. Since 2023, China has issued rules on generative AI services, deep synthesis and algorithmic recommendation; in 2025 the Anti-Unfair Competition Law was revised; and judicial practice has begun to address AI-generated images, algorithmic recommendation and platform filtering duties.
Against that background, the old defence is worth revisiting. Its value is not that the lawyers once said certain things, but whether those propositions have survived the subsequent evolution of legislation, adjudication and scholarship. This article therefore removes the defence from its procedural setting and evaluates it within the coordinates of AI law in 2026: which arguments have been confirmed by history, which remain contested, and which questions still await a better doctrinal framework?
II. Case Background and Reconstruction of the Defence
1. The Technology and Commercial Scenario
The activity in dispute was part of the live quiz business model popular in China between 2018 and 2019. The claimant streamed quiz sessions through its short-video and news apps, attracted users with substantial prize pools, and monetised traffic through app downloads, advertising exposure and user engagement. The defendant’s live quiz assistant was a function within its search app. When users opened it in a floating window, the assistant used OCR to recognise questions displayed on the user’s screen, applied semantic analysis, searched through the search engine, and displayed reference answers. It did not answer on behalf of the user; final selection remained with the user.
Two facts were critical. First, the assistant did not intrude into the claimant’s servers or intercept data streams. It recognised text appearing on the user’s screen. Functionally, this was equivalent to a human reading the question with the naked eye, except that machine vision replaced biological vision. Secondly, it supplied reference answers rather than substitute answers. The user still had to complete the full advertising loop expected by the claimant: downloading the app, entering the live room, viewing advertisements, sharing for revival cards and making the final answer selection.
2. The Claimant’s Theory of Harm
The claimant’s pleaded harm had three elements. It said the assistant increased the number of users answering correctly, thereby undermining the fairness of the activity. It said more winners reduced the average prize per user and forced the claimant to add prizes. It also characterised assistant users as opportunistic traffic hunters who would not remain in the claimant’s app, causing wasted acquisition cost and loss of user-retention interests. On this basis, the claimant alleged that the assistant obstructed the normal operation of its online product and service under the internet unfair-competition provision then in force.
3. The Seven Pillars of the Defence
The defence did not rely on a conventional assertion that the defendant was merely a lawful operator without malice. It reconstructed the legal perception of the case through seven pillars.
- The nature of AI. The assistant was an advanced search tool. OCR, semantic understanding and big-data retrieval were functionally homologous with the human process of reading, thinking and searching. No intrusion or destruction occurred.
- Protection of maverick competitors. Borrowing from the concept of the maverick in US horizontal merger analysis and from Schumpeter’s creative destruction, the defence argued that disruptive AI products should receive special protection under competition law rather than be suppressed through conservative claims of business autonomy.
- Human-machine augmented intelligence as national strategy. The State Council’s New Generation Artificial Intelligence Development Plan identified human-machine collaborative augmented intelligence as a strategic direction. Concrete applications that assist humans should receive positive judicial evaluation.
- Empirical market testing. The behavioural injunction imposed during the litigation created a natural market test. Comparing acquisition costs, pass rates and prize distribution before and after the assistant went offline empirically challenged the claimant’s zero-sum harm narrative.
- Critique of business autonomy. Business autonomy is a remnant concept from the planned-economy era and differs from competitive interests in modern competition law. Treating it as a universal right in unfair-competition litigation is a conceptual substitution.
- Restoration of the activity’s true nature. The activity was promotional, not an examination. It lacked graded assessment, disciplinary constraints and objective evaluation. The language of cheating was therefore misplaced.
- Community of commercial destiny. Data showed that the defendant’s tool had a positive effect on the claimant’s user acquisition and acquisition costs. The parties were not in a zero-sum game; the claimant received traffic benefits from the defendant.
These pillars formed an integrated system. The first and sixth established the factual nature of the technology and activity. The second and third introduced a new evaluative framework. The fourth severed the alleged harm through data. The fifth and seventh dismantled the logic of “business autonomy plus zero-sum loss” on which the first-instance judgment had relied.
III. Doctrinal Deconstruction of the Seven Pillars
1. The Nature of AI: From the Turing Test to the Boundary Between Reading and Hijacking
The central question was whether AI-assisted answering was legally different from human-assisted answering. Functionally, the assistant performed the same sequence that a skilled human user might perform: read the question, understand its semantics, search for information and select a likely answer. The defence used this functional homology to argue that the legal assessment should not turn on whether cognition is biological or computational. If a user may lawfully read, search and think, an AI tool that assists those same acts should not be treated as a system intrusion merely because it is faster.
The boundary is important. A tool that bypasses access controls, manipulates the claimant’s server, automatically answers without user participation or falsifies interaction data may be unlawful. The assistant did none of these. It operated at the user interface level and enhanced user cognition. The proper legal category was therefore assistance, not hijacking.
2. Maverick Competition and Creative Destruction
Competition law should not protect incumbents from the discomfort caused by disruptive tools. A maverick competitor is valuable precisely because it disturbs established business assumptions and forces market participants to adapt. AI assistance is a form of Schumpeterian creative destruction. It may reduce the value of old attention-harvesting designs, but that does not make it unfair. The law should ask whether competition is distorted by improper means, not whether the incumbent’s business model has become less comfortable.
3. Human-Machine Augmented Intelligence
China’s AI policy has long supported human-machine collaboration. The assistant was not an autonomous agent replacing the user; it was an augmentation of the user’s ability to process information. This distinction remains central in 2026. Many AI legal problems arise because courts treat AI as either an independent wrongdoer or a passive tool. The more precise view is to examine how the AI system allocates cognition, decision-making and control between machine and human.
4. Empirical Market Testing
The injunction during the case created an unintended empirical experiment. If the claimant’s theory were right, taking the assistant offline should have improved its acquisition efficiency and prize economics. The data did not support that narrative. The defence therefore shifted the court from moralised language about cheating to measurable market effects. This method remains valuable for modern platform disputes, where alleged harm is often asserted rhetorically but can be tested through traffic, retention, conversion and cost data.
5. Business Autonomy and Competitive Interests
Business autonomy should not be inflated into a general exclusionary right. In unfair-competition law, the protected interest is not an operator’s freedom to maintain any chosen business model against technological change. The protected interest is fair competition. A business model that depends on users not using lawful tools cannot convert user autonomy into the claimant’s exclusive legal interest.
6. Promotional Activity, Not Examination
The claimant’s use of the language of cheating was rhetorically effective but analytically misleading. An examination exists to evaluate ability through graded and disciplined procedures. The live quiz activity existed to attract traffic, expose users to advertising and increase app engagement. In that setting, the user’s use of a search or AI assistant is closer to information consumption than examination fraud. The legal assessment should follow the commercial nature of the activity.
7. Community of Commercial Destiny
The most direct answer to the claimant’s zero-sum theory was that the assistant could increase participation, traffic and advertising exposure. The claimant and defendant were not necessarily competitors locked in a destructive relationship. They were participants in a broader attention economy, and the defendant’s tool may have expanded the claimant’s audience. This anticipates later AI-platform questions: not every external AI layer that changes user behaviour causes legal harm; some create complementary value.
IV. Six Years Later: Legislative, Judicial and Scholarly Confirmation
From 2020 to 2026, Chinese AI law and competition law moved towards many of the assumptions embedded in the defence. Regulatory documents on generative AI, deep synthesis and algorithmic recommendation increasingly distinguish between tool provision, content generation, platform distribution and user decision-making. The 2025 revision of the Anti-Unfair Competition Law confirms that internet unfair competition must be assessed through the actual interference with network products or services, not through abstract discomfort with technological change.
Judicial practice concerning AI-generated images, platform filtering and data rights has also become more granular. Courts now ask what the technology does, where control lies, what evidence of human input exists, and whether the defendant’s conduct crosses from assistance into substitution, misappropriation or obstruction. These questions were already latent in the live quiz assistant case.
The case also anticipated the rise of empirical evidence in platform disputes. Modern litigation increasingly requires traffic analysis, conversion data, A/B comparisons, user-path reconstruction and technical logs. Legal characterisation without economic evidence is often insufficient.
V. Academic Value of the Defence Strategy
The academic value of the case lies in its early articulation of a competition-law method for AI tools. It rejected three common shortcuts: treating all AI intervention as unfair disruption, treating business autonomy as an absolute shield for incumbents, and treating user assistance as equivalent to system interference. In their place, it proposed a functional, empirical and competition-neutral analysis.
First, it placed AI within the law of human assistance and augmented intelligence. Secondly, it introduced the idea that AI tools may be maverick competitors deserving protection because they increase consumer welfare and challenge inefficient incumbency. Thirdly, it insisted that harm be tested, not assumed. Fourthly, it separated technological acceleration from unlawful intrusion. These are still central distinctions in AI law.
VI. Remaining Problems and Future Directions
The retrospective does not mean that all AI-assisted tools should be lawful. Future disputes will require sharper boundaries. A tool that automates final user decisions, bypasses technical protection, fabricates engagement, scrapes non-public data or impersonates a platform environment may require a different result. The legal task is therefore to distinguish enhancement from replacement, assistance from circumvention, interoperability from free-riding, and competition from sabotage.
Future doctrine should develop a four-factor inquiry: the location of the AI tool in the technical chain; the degree of user control; the effect on the claimant’s system and data; and the empirically demonstrable market effect. This would allow courts to protect innovation without tolerating genuinely predatory or obstructive conduct.
Conclusion
The live quiz assistant case was, at the time, an individual dispute. In the longer history of AI law, it was an early local response by Chinese litigation practice to a problem that later became global: how should law evaluate AI tools that assist users in third-party digital environments? The answer proposed by the defence remains useful. The law should not punish a tool merely because it makes users smarter, faster or better informed. It should intervene only where the tool crosses the line into unlawful interference, deception, misappropriation or demonstrable competitive harm.