Hong Kong broker let clients route 10 million futures trades through untested software

A Hong Kong futures broker allowed clients to connect their own trading software to the firm's order-entry infrastructure without first testing or conducting due diligence on those systems, the Securities and Futures Commission found. The failure involved systems used by 160 clients over a period approaching two years, alongside unexplained deposit patterns in eight accounts and 176 potentially suspicious near-simultaneous order pairs. The regulator fined Zheng Da International Financial Holding Limited HK$7 million on 29 September 2026 and suspended responsible officer Zhong Hao for seven months. The findings concern deficient supervision and monitoring; they do not establish that all 160 clients laundered money or that the 176 order instances were criminal wash trades.[1][2]

The case exposes a gap that an ordinary description of electronic trading rarely captures. A broker can control the exchange connection and its internal trading server while clients control the software generating orders upstream. If the broker does not understand that client-supplied software, it may struggle to identify who is sending orders, why the orders arrive in unusual sequences, and whether alerts about deposits or trading are credible. The SFC found that Zheng Da had not conducted due diligence or testing on the customer-supplied systems, commonly abbreviated CSSs, used by the 160 clients during the relevant period, from 1 December 2021 to 30 September 2023.[1][2]

The original contribution of this investigation is to reconstruct three layers of control from the disciplinary record: the untested technology access, the unexamined source-of-funds patterns and the absence of effective trading surveillance. These layers are different, but they can reinforce one another. None proves a crime by a particular customer. Together they explain why regulators do not regard a brokerage account as adequately supervised merely because trades pass through a licensed firm's server before reaching an exchange.[1][3]

The interface is also a risk perimeter

Customer-supplied systems are programs developed or designated by clients that allow electronic trading, including through mobile or internet channels. The SFC explained that the systems in Zheng Da's case connected through application programming interfaces to the broker-supplied system, or BSS, used for exchange access. An API is not itself suspicious; it is a standard method of connecting applications. Its significance is that it delegates some workflow choices to software outside the firm's direct development and testing process.[1]

The broker remains responsible for applicable licensing and supervisory requirements even when the customer creates the algorithm. Testing and due diligence can help determine what functions the software performs, whether an authorised person operates it, how orders are attributed, whether market access can be shared, and whether there are safeguards against unusual or abusive patterns. Those are reasonable control objectives, not claims that every untested client application can enter an order without controls at the exchange. The SFC's finding was that Zheng Da lacked adequate knowledge of the client systems to assess relevant risks properly.[1][3]

A customer might use automated software for legitimate hedging or market making. A futures broker could therefore have excellent commercial reasons to support APIs. The difficulty is proving that its technology permissions, monitoring and customer records continue to match the risks of automated trading. For example, an owner might authorise a legitimate third-party developer to maintain a model, while an unauthorised person could misuse access credentials. These possibilities explain the need for assessment; they are not findings that unauthorised access occurred in this case.[1]

The SFC said the deficiencies exposed the firm to risks including unlicensed activity, money laundering, nominee account arrangements and unauthorised account access. That is language about *exposure to risk*. The regulator did not, in the cited announcement, state that each of those outcomes had actually occurred. Reporting them as established criminal conduct would materially change the determination. An effective broker review would separately inventory the client, account owner, vendor, API user, connected system and ultimate trading authority.[1]

Deposit evidence should have triggered questions

In eight client accounts, the SFC found that the size and frequency of deposits did not fit the financial profiles clients had declared in account-opening documents. This is an important but bounded conclusion. A customer may legitimately receive an inheritance, sell a business or change employment and then transact in amounts above an earlier profile. The compliance obligation is not to presume that every change is illicit. It is to investigate the discrepancy, document the explanation, assess supporting evidence and update the customer risk position where warranted.[1]

Zheng Da asserted that Zhong made enquiries of selected clients about their financial circumstances and deposit activity. The regulator said there were no records supporting those enquiries. The absence of records makes it difficult to demonstrate that concerns were satisfactorily resolved, even if an individual recalls a conversation. A supervisor cannot reconstruct why a risk was dismissed from an undocumented assurance. The regulatory finding should not be exaggerated into proof that no staff member ever spoke with a client; it establishes failure to demonstrate proper enquiries and an inadequate monitoring framework.[1]

Large receipts alone do not equal proceeds of crime. A properly designed rule might use cumulative deposits relative to declared income, recent turnover, business purpose, countries of origin and linked parties. It should also produce a review trail: the alert, source information, customer explanation, assessment and outcome. Zheng Da's case points to a missing documentary chain rather than a published bank-by-bank demonstration that the cash itself was unlawful. No amount should be attributed to criminal proceeds on the basis of these discrepancies.[1][3]

Same-second orders are a surveillance signal, not a verdict

The SFC identified 176 instances in which the same client placed buy and sell orders in the same futures contract, at the same price and within the same second. Such combinations can indicate potential fictitious or manipulative activity and therefore merit review. But the pattern is only a lead. In a liquid futures market, a professional trader can send opposing instructions while adjusting an exposure or operating several legitimate strategies. Whether those orders trade against one another, were intended to create false activity or were matched with independent counterparties must be established from execution and order-book data.[1]

The regulator's criticism was that Zheng Da lacked effective ongoing monitoring capable of detecting and assessing these patterns. A system that records an order without recognising a paired or repeated pattern may meet basic transaction-booking needs while missing market-integrity indicators. Conversely, an alert alone cannot establish guilt; the value comes from a review that joins order identifiers, client identity, beneficial owner, execution time, contract, price and account explanation. The 176 count is the number of flagged patterns identified in the enforcement record, not 176 judicially proven wash trades.[1][3]

This difference is particularly important when traders use APIs and algorithmic strategies. Electronic systems can produce far more orders than a human dealer could type manually, and a simple threshold may flood an investigator with false alerts. Surveillance can be designed around ordering sequences, self-match probability, changes of participant identity and net economic position. The objective is not to block every client who operates on both sides of a market; it is to give compliance staff the information needed to evaluate a suspicious pattern in context.[1]

The supervision problem was not confined to one desk

Zhong's responsibilities extended beyond a nominal futures dealing licence. The SFC's notice identified his role as a responsible officer for Type 2 dealing and Type 5 advising on futures contracts, together with senior-management responsibilities covering oversight, anti-money-laundering, compliance and technology during relevant periods. Those responsibilities matter because the failures sat across systems governance, transaction monitoring and records. A product team could not cure them merely by handing the same alerts to a different person without improving the underlying controls.[1]

The regulator suspended his licence from 28 September 2026 to 27 April 2027, a seven-month interval, while imposing a HK$7 million fine on the company. One penalty is a monetary sanction against the legal firm, the other is a time-limited licensing restriction on a particular individual. They should not be merged into a claim that Zhong personally paid HK$7 million. The disciplinary records described compliance failures attributable to his supervisory duties, not a public conviction for money laundering.[1][2]

The regulator also recorded mitigating and aggravating circumstances. Zheng Da and Zhong cooperated in resolving concerns and otherwise had clean disciplinary records. But the SFC stressed that the company had continued permitting customer-supplied software despite a series of earlier enforcement actions involving similar issues across other licensees. Zheng Da stopped accepting new CSS applications and disabled existing CSS connections in August 2025, according to the enforcement announcement. That change came after the 2021–23 period under examination and before the September 2026 discipline. The dates matter when describing present risk.[1]

The enforcement body referred to earlier cases involving multiple futures firms. Those references demonstrate an industry-wide regulatory message about customer-supplied systems; they do not establish that the companies were part of the same criminal network. Similar compliance failures can arise independently where firms use comparable systems or business models. A proper network allegation would require evidence of shared beneficial ownership, counterparties or coordinated conduct. None should be inferred merely because the firms appear in a regulator's notes.[1][3]

The three unjoined records reveal the operational lesson

The most consequential reconstruction links a technology question to two data streams: the clients' ability to send orders through untested systems; deposits whose scale did not fit known financial profiles; and suspicious-looking order pairs that surveillance did not effectively identify. Each layer contains only part of the necessary evidence. A deposit review might not see the trading algorithm; a trading desk might not have the customer's income profile; a technology team might know the API key but not its beneficial user. A supervisory program needs a way to join them.[1]

That linkage offers an original benchmark for other futures brokers. The useful control is not simply “approve trading software” once. It includes a maintained inventory of client-connected applications, periodic revalidation after software changes, permission and credential records, integrated transaction monitoring, documented escalation and tests of whether alerts actually reach a responsible reviewer. These are proposed compliance tests derived from the SFC's findings. This investigation did not access Zheng Da's current platform or perform a security assessment of its API.[1][3]

Nor should the counts be combined into a false rate. The 160 is a count of clients using customer-supplied systems; eight is a count of accounts with deposit-profile discrepancies; 176 is a count of order-pattern instances. The records do not establish one uniform population or show whether each deposit account belonged to the CSS cohort. Dividing 176 by 160 would not produce a meaningful probability of wrongdoing, and adding the counts would not yield a number of affected customers. Their significance is the coexistence of distinct unmanaged risk channels.[1]

What remains open

A final customer-harm assessment would need evidence of actual suspicious transactions, whether exchanges cancelled or adjusted relevant executions, any affected counterparties and customer assets, and what later monitoring detected after the August 2025 system changes. The published disciplinary announcement supplies none of those as a quantified loss schedule. It also does not establish that the two named enforcement subjects responded to fresh questions for this article. No outreach has been made, and any request for further explanation should be sent before publishing allegations beyond the SFC's own findings.[1][2]

The defensible conclusion is significant without embellishment. A Hong Kong licensed futures broker was disciplined for failing to supervise 160 customers' trading software and for deficiencies in monitoring unusual deposits and order patterns; its responsible officer was suspended for supervisory shortcomings. The unresolved question is how many trading firms elsewhere can demonstrate that their client-controlled APIs, funding alerts and order surveillance actually produce one coherent, reviewable picture. That is a test worth conducting before a suspicious trade becomes a proven loss.[1][3]

Report an error or submit an update