A Singapore Contract, Beijing’s Red Lines

Corporate self-censorship to preserve Chinese market access is old news: a boycott, an apology, a firing. Roy Jones, the Marriott worker fired in Omaha for liking a tweet from a Tibetan group, told VOA he had been given no guidance about China. Pressure like that reaches workers sideways and late, through an employer already in crisis. Polly’s screen reverses the sequence: the restriction is written down, presented before work begins and accepted by a click, in unmistakably PRC political vocabulary. The platform form makes that reversal cheap. A company can incorporate in one jurisdiction, sell in a second, recruit contractors across many more, and govern them all through the same click-through terms. If Nvidia compelled TSMC to accept comparable political terms about the names of the Gulf of Mexico or Lake Ontario, the extraterritorial implications would be hard to miss. The method is certainly no less consequential because the pressure originates in Beijing. It shows how the corrosive labor logic of “the platform” becomes a vehicle for political control.

On paper, the teachers are independent contractors. In practice, Polly sets the courseware and ratings and controls the booking flow and pay tiers that decide who gets students and what they earn. Arrangements like this have come under growing scrutiny for the tension between nominal worker autonomy and opaque systems that determine access to work, pay and visibility. Polly also decides when an account ends. A political rule needs no enforcement apparatus of its own. It rides the one that already decides who teaches and who stops earning.

A 2024 controversy involving another Singapore-based online teaching platform, Lingostar, shows what this mechanism can look like in practice. By Lingostar’s own account, a new British teacher told a Chinese student that Taiwan was a country, not part of China, and made a vulgar gesture. In a statement reported by Modern Express, Lingostar said it upheld the one-China principle and had dismissed the teacher for deliberately breaking its ban on political remarks in class. Later, the China Private Education Association announced that, after checking with the relevant authorities, it had found the operation unlicensed and non-compliant, and that its apps had been removed pending investigation.

That scrutiny collided with an already restrictive environment for online tutoring. China’s 2021 “double reduction” directive, part of a broader crackdown on private tutoring for compulsory-education students, prohibits training institutions from engaging foreign personnel located outside China to provide training. Its reach therefore depends on which students a platform serves. Polly recruits teachers for students aged three to eighteen, a range that includes compulsory-education students. What matters is how it is administered. Reuters reported in 2024 that tutoring businesses were again operating more openly despite no formal reversal of the crackdown. Formal restriction and regulatory tolerance can coexist, leaving operators dependent on discretion and creating a strong incentive for self-policing. This is an arrangement very typical of the PRC.

Sequence matters. Complaints drew scrutiny, the scrutiny found the operation already non-compliant, and the apps came down. The Taiwan remark did not itself legally cause the delisting any more than the platform’s prior non-compliance with the “double reduction” directive did. One started the process that produced it; the other was the conveniently pre-existing condition it ran into. A contractor can generate regulatory exposure for a platform, and the platform has contractual means to push it back. Control stays centralized while risk travels outward, across national borders. As I argued in a previous piece, self-policing of that kind does most of the work; direct enforcement is rarely needed. Again, all too typical of the PRC, and worth taking note of as such.

The corporate structure disperses the same way the risk does. Mainland outlets named Chinese authorized companies behind Lingostar in Shenzhen, Beijing and Tianjin; Taiwanese coverage citing mainland reporting described the Shenzhen entity as a shell. As of September 2026, Lingostar’s own recruitment portal was carrying Polly-branded vacancies, one listing requiring teachers to use Classpod. On Tencent’s app store in the PRC, Polly’s Android app is listed as developed and operated by 北京有志于道技术科技有限公司 (京ICP备2024095633号-5A). None of that establishes common ownership. It does mean that “based in Singapore” describes the contracting entity more clearly than the operation as a whole.

Before the 2021 “double reduction” directive, a Hong Kong contract could still serve as an offshore wrapper for this model. After 2021, the incentives changed, just as Beijing’s ambitions for extraterritorial control were becoming more explicit. Those guidelines went further than trial in absentia: in the gravest cases they allow capital punishment for alleged Taiwan-independence separatism. The ethnic unity law took effect in July 2026, and Amnesty International warns that the authorities’ stated intention to apply parts of it beyond China’s borders could strengthen existing practices of transnational repression. What enforcement abroad would look like is unspecified, an omission hard to read in isolation from documented PRC transnational repression, including harassment of critics overseas and alleged overseas police operations. Taiwan’s Mainland Affairs Council warns it could institutionalize transnational repression against Taiwanese citizens; the European Parliament calls its extraterritorial provisions transnational repression.

None of that reached the Lingostar teacher, and no evidence connects those measures to Polly’s contract. The teacher’s words produced commercial and regulatory consequences, not criminal ones. The point is convergence, not coordination. The platform model is what makes it available. Beijing can push its red lines outward while private companies already hold infrastructure capable of reproducing them, which lets the state expand its claims without administering the consequences. That is also what plausible deniability looks like from the outside: no single document proves a command chain, yet the incentives, rules, and enforcement pressures line up. Centralized control presents itself as an open market, and responsibility lands on whoever sits furthest from the center.

Whatever political logic lies behind the restriction, the legal shell is Singaporean. Polly’s teacher terms choose Singapore law and give Singapore courts exclusive jurisdiction. If the screen’s threat of legal action means anything, the question is whether Singaporean private law can be made to carry the economic consequences of a political boundary drawn in Beijing.

In practice, the cheapest enforcement never reaches a court. Platform work is governed through allocation, ratings, suspension and deactivation, and Polly already controls those levers. Recovery is slower. Polly’s terms contemplate proceedings to recover costs arising from a breach, further legal action and liability for damage to the company or to third parties. Singapore damages law still requires an identifiable loss, causation and remoteness. Lingostar suggests what such a chain might look like; whether it would survive in court is another matter. A contractor earning the fifteen to twenty-four dollars an hour Polly advertises does not need a verdict to be disciplined by the threat.

The terms also contemplate disclosure to law-enforcement authorities where the company reasonably thinks it necessary. The clause is generic: it names no jurisdiction or authority, it appears in agreements with no political content, and there is no evidence Polly has ever used it over Taiwan, Tibet or Xinjiang. But a contract that forbids discussion of five PRC red lines also reserves the right to hand information to authorities it does not name. That proves no reporting pipeline. The worker is asked to accept the uncertainty anyway.

One clause makes the screen harder to dismiss as advisory: Polly’s terms state that notices, warnings and statements about particular services become part of the agreement. Singapore courts can refuse to enforce contracts that offend statute or established common-law public policy, but in Ochroid Trading v Chua Siok Lui (2018) the Court of Appeal warned that those categories are not closed and will not readily be expanded. So the legal limits stay unsettled. The practical asymmetry does not: the platform states the rule now, and the worker would spend money later to discover its limits.

That imbalance is political, not natural, and Taiwan has said so in statute. Its 2026 Delivery Workers’ Rights Protection and Delivery Platform Management Act does not cover online teachers, but it starts from a premise platform capitalism prefers to leave unstated: a contract between a platform and an ostensibly independent worker is not necessarily the product of equal bargaining power. The Ministry of Labor justified it by pointing to platform control over dispatch, pay, and contract drafting. Article 4 imposes mandatory and prohibited terms, voids nonconforming provisions, and requires worker consent for important changes; Article 7 limits termination to serious breaches of enumerated laws or the contract. The statute is narrow and sector-specific; the precedent is not. It reasserts public authority over private rule-making, treating platform contracts as instruments of power rather than neutral descriptions of exchange.

Which is what makes Polly’s screen larger than a China-policy curiosity. A teacher in Taiwan can sign with a Singapore entity, teach children recruited through a Simplified Chinese site and app stores in the PRC, and accept a prohibition on discussing Taiwan backed by termination, legal action, and damages. No censor has to cross a border. No PRC statute has to be served on the worker. The political demand arrives as a term of service, enforced through the same infrastructure that allocates classes, calculates pay, and decides who stays on the platform.

Beijing’s red lines are the most visible part of this story. They are not the deepest. Underneath is a system that centralizes private power while fragmenting public accountability. Workers are independent when risk is assigned to them, subordinate when the rules are written. Markets stay “free” while access is governed by systems no worker can meaningfully negotiate. Once platforms become cross-border infrastructure, they carry more than prices, ratings, and labor practices. They can carry political sovereignty. Political censorship makes that capacity unusually easy to see. The machinery was already there.