The Trust Was Never the Shield: China's Offshore Tax Sweep Reaches Crypto's Family Offices

The most expensive word in cross-border tax enforcement right now is not "evasion." It is "waiver."

On 22 October 2023, a ninety-day clock stopped. The window had opened on 24 July 2023, when the Ministry of Finance and the State Taxation Administration jointly issued what the first-phase analysis of this file calls Announcement No. 21 of 2023 — a departmental normative document aimed squarely at the offshore trust beneficiaries that China's tax system had spent two decades politely not looking at. The deal on the table was simple and, in the language of enforcement, unusually generous: come forward, declare the distributions you have been receiving from structures in the BVI, the Cayman Islands, Jersey, Singapore, Hong Kong, pay the individual income tax you owe, and the late-payment surcharge — five basis points per day, roughly 18.25% annualized — is waived.

What the window did not waive was the fine. And that distinction, buried in a document that most of the people it targets will never read in full, is the single most consequential line in the entire exercise.

I want to be precise about my sourcing here, because precision is the whole job. I have not obtained the full text of Announcement No. 21. Its formal title, its clause numbering, its income-classification rules — all unverified. Everything I say about the announcement's content is drawn from the first-phase breakdown and flagged as such. What I can verify, article and clause and subparagraph, is the statutory scaffolding it sits on: the Individual Income Tax Law as amended in 2018 and effective 1 January 2019, its Implementing Regulations under State Council Order No. 707, the Tax Collection and Administration Law and its detailed rules, and Article 201 of the Criminal Law. Those are the load-bearing walls. The announcement is the wallpaper.

And the wallpaper, in this case, is hanging over a room that crypto built.

Because here is what changed between the last enforcement cycle and this one: the assets sitting inside those offshore trusts are no longer just Shanghai apartments and Hong Kong brokerage accounts. They are token positions. They are SAFTs and vested equity in offshore issuers. They are stablecoin treasuries, staking yield streams, NFT royalty flows, and LP positions in protocols whose governance tokens were airdropped to wallets that nobody has ever bothered to document. The offshore trust industry spent a decade selling a product to exactly this cohort — the newly liquid crypto founder, the early miner, the DeFi native who cashed out in 2021 and wanted a structure. That product was never really a tax product. It was a confidentiality product. And confidentiality, in the age of automatic exchange, is a depreciating asset with a known maturity date.

So let me tell you what the ninety days actually were, what the document actually says, and where the real exposure sits — not in the headline tax bill, but in a burden-of-proof inversion that most of the market has not yet priced.

Context: a trust is not a taxpayer, and that changes everything

Start with the premise that generates most of the professional malpractice in this space. Under Chinese tax law, a trust is not a taxable entity. It has no independent civil subject status for tax purposes. It does not hold a taxpayer identification number. It does not file. The taxpayer is the individual — the settlor, the beneficiary, the person with effective control — and if that individual is a Chinese tax resident, the exposure is on worldwide income, full stop. Article 1 of the Individual Income Tax Law is not ambiguous about this. Resident individuals pay tax on income sourced inside and outside China. Article 2 enumerates nine categories of taxable income. Article 3 sets the rates. Article 8 provides three anti-avoidance rules that have been sitting there, largely dormant, since 2019.

The structure, in other words, has never been the taxpayer. The person behind the structure has always been the taxpayer. Every family office that told a client "the trust owns it, you don't" was describing a legal fiction that works beautifully against creditors and catastrophically badly against a revenue authority that has decided to look through.

And they have decided to look through. The regulatory architecture that makes this possible was assembled quietly and in pieces: China's CRS implementation, with first exchanges in September 2018; the consolidation of tax data under the Golden Tax Phase IV system; the linkage of bank accounts, real estate registries, and market-supervision equity-change records. Each of those is unremarkable alone. Stacked, they form a triangulation engine. The 90-day window was not a legislative courtesy extended to a confused population. It was a data-matching deadline dressed as an amnesty.

The announcement itself sits low in the hierarchy of norms. It is a departmental normative document — below administrative regulations, below the law it purports to interpret. Its legitimate function is interpretive and executory. It cannot create a tax base. It cannot grant or revoke a liability. That constraint matters enormously here, because the announcement appears to have done the one thing it structurally should not have needed to do and did not do: it left the classification question open.

Article 2's nine categories are not a menu you can pick from casually. The same distribution from the same trust, received by the same person, could plausibly be characterized as dividend and interest income at 20%, as property transfer income at 20%, or as other income at 20%. Identical headline rate. Wildly different consequences. The tax basis differs. The allowable deductions differ. The timing of the liability — the moment the obligation crystallizes — differs. Whether a loss in one year can offset a gain in another differs. Whether the trustee's distribution of principal counts as a return of capital or a taxable event differs.

When a statute leaves classification to the front line, the front line becomes the legislature. That is the structural reality here, and it is why two taxpayers with economically identical facts can receive two different assessments depending on which tax bureau's risk-management division opens the file.

Now add crypto to that vacuum. A trust holding a token position generates at least five distinct economic events that Chinese individual income tax law has no settled treatment for: the airdrop (income on receipt, or zero-basis acquisition?), the staking reward (other income, or a service fee, or a dividend-equivalent?), the NFT royalty (property transfer, or licensing income?), the LP fee (business income requiring registration, or incidental investment income?), and the token-for-token swap (a realization event, or not?). None of these are answered. All of them are now in scope, because the trust that holds them has been dragged into the light.

I have spent the better part of four years mapping trust-held wallet clusters for clients who wanted to know how visible they were. The answer, in almost every case, was more visible than they believed and less visible than the revenue authority assumed. That gap — between belief and reality on both sides — is where the entire next phase of this story lives.

Core: six places where the exposure actually sits

One. The classification vacuum is not neutral. It runs against you.

When the statute is silent and the announcement does not fill the silence, the default is not "no tax." The default is administrative discretion, and administrative discretion in a举证-reversed environment — and I want to state this plainly, because it is the mechanism that determines outcomes — resolves against the taxpayer almost every time.

Here is the mechanics of it. If a distribution cannot be shown to be a return of principal, it will be treated as income. If it cannot be shown to be capital gain with a documented basis, it will be treated as ordinary income at the full amount, no deduction. If it cannot be shown to be a specific enumerated category, it will be swept into "other income" — Article 2's residual bucket — which carries 20% with essentially no deductions and no loss offset whatsoever. "Other income" is not a category. It is a default judgment.

For crypto specifically, this creates a perverse incentive structure that the market has not internalized. A founder who took tokens as compensation in 2017, put them in a BVI trust in 2019, and has been receiving distributions since 2021 is now facing a question she cannot answer with documents she does not have: what was the acquisition cost of the tokens, in renminbi, at the time they entered the trust? If the answer is "I don't know," the answer the tax authority writes down is zero. Twenty percent of the gross distribution, no basis offset, no argument.

Two. There is no Form 3520 equivalent, and that absence is the most underrated risk in the document.

The United States solved this problem in 1996 with a reporting form. Form 3520 forces disclosure of the trust's existence, its beneficiaries, its distributions, and its trustees, with penalties reaching 25% of the unreported value for failure to file. It is brutal, it is effective, and — critically — it creates a structured data stream that the revenue authority can match against other sources without ever reading a narrative.

China has not built that. The first-phase analysis is unambiguous on this point: the announcement establishes a tax liability without establishing a reporting vehicle, and without an independent penalty for non-reporting. On its face, that looks like a gap that favors the taxpayer.

It is the opposite. Here is why, and this is the part I would put in bold on a whiteboard in front of every family office in Singapore and Hong Kong.

When there is no structured form, the taxpayer's own narrative becomes the evidentiary record — and a narrative is infinitely easier to impeach than a form. A form has fields. Fields are checkable. A narrative is prose, and prose can be contradicted, and contradiction is the gateway to an intent finding. In a structured-disclosure regime, a mistake is a mistake. In a narrative regime, a mistake is a data point in a pattern, and a pattern is how you get from the failure-to-file provision of the Tax Collection and Administration Law to the evasion provision — from an administrative penalty to a criminal referral.

I have watched this dynamic play out in other contexts. When you hand an examiner a stack of documents with no organizing schema, you are not being transparent. You are being searchable. The absence of a form does not reduce your exposure; it transfers control of the frame to the person assessing you.

Three. CRS is weaker than the market thinks and stronger than the market hopes.

There is a widespread belief in the crypto offshore world that CRS cannot see through a discretionary trust, and therefore the trust remains a functioning opacity device. That belief is half right, and the half that is wrong is the half that gets people indicted.

CRS exchanges financial account information. The reporting obligation sits on the financial institution, and the account holder of record for a trust account is the trustee — or, depending on the jurisdiction's implementation, the trust itself as a passive non-financial entity. The trust's beneficiaries are reportable only where the account is held by a passive NFE and the jurisdiction's implementation requires look-through to controlling persons. That look-through is not uniform. For a discretionary trust — where no beneficiary has a fixed, enforceable entitlement until the trustee exercises discretion — the identification of a "controlling person" is genuinely contested in several implementing jurisdictions, and the reporting quality reflects that contest.

Add the second layer: BVI VISTA trusts and Cayman STAR trusts deliberately strip the trustee of the duty to monitor and intervene in the underlying company. That was designed for asset protection. It has the incidental effect of muddying the control analysis for reporting purposes. A trustee who genuinely has no duty to monitor is a trustee who can, in good faith, report a thinner controlling-person set than a Chinese examiner would expect.

So there is a real information asymmetry window. It is not fictional. It exists.

And it is closing on a schedule that anyone holding these structures should already have in their calendar.

The OECD published the Crypto-Asset Reporting Framework in 2023, and it is the thing that ends this argument. CARF extends the automatic-exchange architecture to crypto-asset service providers — exchanges, brokers, and, in many implementations, certain dealers. The European Union's DAC8 transposes it, adopted in 2023 and applying from 1 January 2026. The first wave of roughly four dozen jurisdictions committed to commencing exchanges by 2027. China was not, as of my last review, on the early-adopter list. That buys time. It does not buy immunity.

The honest framing is this: the CRS gap on discretionary trusts is a two-to-four-year window, and the CARF rollout is the thing that closes it — not by improving trust look-through, but by making the on-ramp and off-ramp of every token position reportable at the exchange level. You do not need to pierce the trust if you can see the wallet's fiat leg. The trust becomes irrelevant the moment the exchange that cashes out the distribution is a reporting institution.

In the void, we found our value in the noise. Here the noise is on-chain, and the value is in the fiat rail.

Four. The clause nobody quoted is the one with the teeth.

Article 8 of the Individual Income Tax Law contains three anti-avoidance rules, and the commentary on this announcement has, in my reading, focused on the wrong two.

The first is the related-party rule — transactions not at arm's length. The third is the general anti-avoidance rule — arrangements lacking reasonable commercial purpose. Both are familiar. Both require the authority to characterize an arrangement, which means both are contestable.

The second is the one that matters: the controlled foreign corporation rule. Where a resident individual controls an enterprise established in a low-tax jurisdiction, and that enterprise does not distribute profits without reasonable business need, the tax authority may deem those profits distributed and tax them to the resident individual.

Read that against the standard offshore trust architecture and watch what happens. The typical structure is not a bare trust holding assets directly. It is a trust holding a BVI or Cayman holding company, which holds the operating entity or the token treasury. The trust does not distribute. The holding company does not distribute. Everyone congratulates themselves on the accumulation strategy.

The CFC rule does not care. If the individual controls the enterprise — and control tests under these rules look to substance, including de facto control through a trustee relationship — and the jurisdiction is low-tax, and there is no reasonable business need for the non-distribution, the profits can be attributed to the individual regardless of whether a single dollar ever leaves the structure.

This is the sentence that kills the "we don't distribute, so there's nothing to tax" defense. It was never a good defense. It has simply never been tested at scale in the individual income tax context. The announcement does not mention it. The announcement did not need to. It is in the statute, it has been in the statute since 2019, and the only thing that has changed is that the enforcement apparatus now has enough data to apply it.

A trust that holds a token treasury through a Cayman wrapper, with no operating substance, no employees, no premises, and no commercial activity beyond holding, is a textbook CFC fact pattern. The reasonable-business-need defense requires evidence of actual business purpose. "Tax efficiency" is not a business purpose. It is the thing the rule exists to defeat.

Five. Waived late fees are not waived penalties, and the criminal chain does not close.

This is the point I would hammer until people stop nodding politely and start panicking appropriately.

The 90-day window waived the late-payment surcharge. Under Article 32 of the Tax Collection and Administration Law, that surcharge accrues at five basis points per day — 0.05%, annualized to about 18.25%. For a taxpayer with a decade of undeclared distributions, that surcharge is often larger than the tax itself. Waiving it is genuinely valuable. It is also, and only, an economic accommodation. The surcharge is compensatory in nature: it is the price of having had the use of the money.

Fines are different. They are punitive. Under Article 63, tax evasion carries a fine of 50% to five times the underpaid tax. Under Article 64(2), failure to file carries the same 50% to five-times range. Under Article 64(1), fabricating a false tax basis carries a fine up to 50,000 renminbi. The announcement, as characterized in the first-phase analysis, waives the surcharge and does not say a word about waiving the fine.

That silence has a consequence that almost nobody in the offshore advisory world is explaining correctly, so let me spell it out.

Article 201(4) of the Criminal Law provides a first-offense exemption from criminal liability for tax evasion — but only where the taxpayer, after being ordered by the tax authority to pay, has paid the tax, paid the late fee, and received an administrative penalty. All three. The exemption is conditioned on the penalty having been imposed and accepted. It is not conditioned on the penalty having been waived.

A taxpayer who used the window to pay tax and escape the surcharge, but who never resolved the penalty question, has an open criminal exposure that they believe is closed. They have done two-thirds of the work and told themselves the job is finished. That is the compliance hallucination at the center of this entire exercise, and it will not surface until the first case where a taxpayer discovers that a closed assessment is not a closed file.

There is a second timing trap layered on top. Article 52 of the Tax Collection and Administration Law sets the limitation period for recovery of underpaid tax at three years, extended to five where the underpayment exceeds 100,000 renminbi. Where the underpayment arises from evasion, resistance, or fraud, there is no limitation period at all. The strategy of sitting still and letting the clock run is therefore not a strategy. It is a bet that your file is characterized as negligence rather than evasion, in a regime where the characterization decision belongs to the examiner and where the narrative — see point two — is the evidence.

Six. How they actually find you, and why it is not the chain.

Here is the part the crypto-native cohort consistently gets wrong, and I say this as someone who has spent years doing the opposite of what they fear.

On-chain analysis is not the primary enforcement vector. It is a corroboration tool. Wallet clustering, exchange deposit-address attribution, bridge-flow tracing — all of it is real, all of it is mature, and all of it is most useful after someone has already told the authority where to look. The chain gives you a graph. It does not give you a name. The name comes from the fiat leg.

The fiat leg is where these cases are actually built. A distribution lands in a Singapore or Hong Kong bank account and gets wired to China to buy property. A token position is sold on an exchange with full KYC and the proceeds settle into an account that is CRS-reportable. A holding company's shares are transferred and the equity-change registration shows up in the market-supervision database. A disgruntled co-founder, a former spouse, a terminated trustee — a whistleblower with a document set.

Each of those is cheap to investigate. Each of them produces a paper trail that predates the tax question entirely. The enforcement sequence described in the first-phase analysis — remind, then coach, then risk-response, then audit — is a funnel that starts at the cheapest evidentiary entry point and works inward. The "reminder" phase is not public education. A public notice is addressed to everyone. A reminder is addressed to a list. The distinction matters because the delivery method determines the moment of knowledge, and the moment of knowledge determines whether a subsequent failure is characterized as negligence or as intent.

And here is the uncomfortable corollary for the crypto cohort: the very structures that made you feel invisible are the ones that generate the most paper. A bare wallet is anonymous. A trust with a trustee, a registered agent, a corporate service provider, a bank account, and a set of audited holding-company accounts is a filing cabinet with your name on the spine.

DeFi was not a bug; it was a feature of chaos. Offshore trust structuring is the same chaos, productized and invoiced at an annual retainer. And productized chaos leaves receipts.

Contrarian: the tax bill is the smallest part of this

Everyone is running the arithmetic. How much tax on a $20 million distribution, at 20%, with what basis, over how many years. That arithmetic is the least interesting number in the file.

Here is the contrarian read, and I will stake my reputation on it: the offshore trust industry did not sell tax avoidance. It sold delay. And delay has now been repriced.

For twenty years, the pitch was that the structure put distance between the individual and the asset — distance from creditors, from disclosure, from the state. That distance had a cost, and the cost was paid in annual fees to trustees and registered agents who provided a service that was 20% legal engineering and 80% administrative patience. The tax outcome was always going to be determined by the residence of the beneficiary, not the address of the trustee. Everyone in the room knew this. The clients did not, because the clients were buying peace of mind and peace of mind is sold, not calculated.

What the 90-day window did was collapse the delay. Not the structure. The delay. The structure still works. It still protects assets from creditors. It still governs succession. It still does everything it was designed to do, except one thing it was never legally able to do, which is change who owes tax on the income.

The second contrarian point is harder, and it will annoy people on both sides.

Onshoring does not reduce your risk. It increases your audit surface.

The instinctive response to a cross-border enforcement wave is to bring the structure home — dissolve the BVI company, distribute in specie, hold the tokens in a domestic entity, become legible. It sounds like the responsible move. It is frequently the highest-risk move available.

Consider what happens when a token position moves from an offshore trust to a domestic holding company. You have just converted a foreign-source question into a domestic-source question. You have created a resident enterprise with a resident tax registration, resident filing obligations, and a resident examiner with jurisdiction. You have created a valuation event at the moment of contribution. You have potentially created a deemed-transfer event on the contribution itself. And you have put a Chinese tax authority in direct, unmediated contact with an asset class whose fair value is disputed, volatile, and — critically — whose cost basis you still cannot document.

The offshore structure at least gave you a jurisdiction question. The onshore structure gives you an audit. Those are not the same risk. One is a conversation between two revenue authorities with a mutual agreement procedure at the end of it. The other is a desk audit with a deadline.

The third thing nobody is saying: the cohort that is actually in trouble is not the whales. It is the middle.

A family with $500 million under management has a Big Four firm, a legal opinion from a magic-circle shop, a voluntary disclosure strategy, and a relationship with a private bank that has already walked them through three of these cycles. They will pay, they will restructure, they will be fine. A holder with $4 million in tokens distributed from a trust set up in 2021 by an introducer who has since left the industry has none of that. They have a WhatsApp thread, a PDF trust deed they have never read, and no idea that the surcharge they just avoided is not the same thing as the fine they have not.

The middle of this market is where the enforcement pain will concentrate, precisely because it is where the cost of compliance exceeds the cost of the tax. That is not a moral observation. It is an arithmetic one, and arithmetic is what enforcement responds to.

I would add one more thing, and it comes from a place that is not usually in these analyses.

Not everyone in an offshore structure is there to avoid tax. In Lagos, in Buenos Aires, in Istanbul, the demand for dollar-denominated savings is not ideological. It is a survival response to currencies that lose a third of their value in a year. When a trader converts naira into stablecoins and parks them somewhere a local bank cannot reach, the driver is not tax arbitrage. It is the entirely rational desire to not watch your savings evaporate. The compliance apparatus, however, does not distinguish between motive and mechanism. It sees a structure and it sees an unreported account. The gap between why people build these things and how the state reads them is where the real human cost of this wave will land — and it is a gap that no announcement, no transition period, and no reporting form has ever addressed.

Takeaway: what to watch, and the one question that matters

Four things, in order of when they will hit.

First, the reporting vehicle. Watch for a supplementary document establishing an offshore trust disclosure form, an offshore income schedule, or a filing protocol for trust distributions. That document will matter more than the announcement itself, because it will define what "compliance" actually looks like and what a failure to comply is measured against.

Second, the first public case. Not a rumor, not a "my accountant says" — a disclosed assessment or an administrative reconsideration decision involving an offshore trust distribution. That case will set the classification precedent, and the classification precedent will determine the basis question, and the basis question is where the money is.

Third, the perimeter expansion. The first-phase analysis flags this and I agree with the trajectory: the enforcement center of gravity moves from trust distributions to undistributed offshore company profits under the CFC rule, then to offshore insurance wrappers and family office structures. The trust is the opening. The corporate layer is the main event.

Fourth, the treaty question. Where a beneficiary is resident in a jurisdiction with a tax treaty, the characterization of a trust distribution as "other income" raises an allocation question, and allocation questions go to mutual agreement procedure. Watch for the first case that gets there. It will tell you whether the framework is being applied with technical rigor or with revenue urgency.

The story isn't in the price. It's in the pulse. And the pulse here is not the 20% rate, not the 18.25% annualized surcharge, not the 90-day window that closed on a Sunday in October.

The pulse is the burden of proof. The document did not create a tax. The tax was always there, sitting in Article 1 of a statute that has applied to worldwide income since 2019. What the document did was announce that the state now intends to collect it — and that the evidence required to contest it will have to come from you.

So here is the question every holder of an offshore structure should be able to answer tonight, without calling anyone, without opening a laptop: what is the acquisition cost, in renminbi, of every asset inside your trust, and where is the document that proves it?

If the answer takes more than ten seconds, you do not have a tax problem. You have an evidentiary one. And the evidentiary one is the one that does not get waived.