Hong Kong's Banks Already Issue Its Cash. A BIS Prototype Shows How Their Digital Money Could Keep Cash's Privacy.

Most of the banknotes in a Hong Kong wallet were not issued by the central bank. They were issued by one of three commercial banks: HSBC, Standard Chartered (Hong Kong) or Bank of China (Hong Kong), each authorised by the government and each redeeming its notes against US dollars paid into the Exchange Fund at a fixed rate of HK$7.80. Hong Kong has run on privately issued, fully backed cash for a very long time.

That arrangement is why the city's digital money experiments look the way they do. In October 2022, the first phase of Project Aurum, run by the BIS Innovation Hub with the Hong Kong Monetary Authority, built a CBDC-backed stablecoin and described it explicitly as the digital equivalent of the note-issuing bank system: issued by a bank, a liability of that bank, and fully backed by central bank money locked in a custodian account.

Four years on, the analogy has a real-world counterpart. On 10 April 2026 the HKMA granted Hong Kong's first two stablecoin issuer licences, to HSBC and to Anchorpoint Financial, a venture backed by Standard Chartered (Hong Kong). Two of the city's three note-issuing banks are now behind its first licensed stablecoins.

The second phase of Aurum, just published, takes up the question the first phase left open: if a bank-distributed digital token is going to replace some share of cash, how much of cash's privacy should it keep, and who should be able to see what? We've focused on the privacy design and its trade-offs, and on how they read across to bank-issued stablecoins, which is our extension of a report that is about central bank money.

What Phase 2 actually tested

One clarification matters before anything else. Phase 2 does not test a bank-issued stablecoin. Its token is digital cash that represents a liability of the central bank. Commercial banks distribute it to customers by locking an equivalent amount of tokenised central bank reserves in a smart-contract custodian account, the same backing mechanism Phase 1 used, so that every coin is fully backed and auditable.

What Phase 2 adds is a privacy layer on top of that architecture, built by the BIS Innovation Hub Hong Kong Centre with the HKMA, the Hong Kong Polytechnic University and ASTRI. The report describes its coins as largely platform-agnostic from a transfer and verification perspective, and it names stablecoins directly once, noting that the same pseudonym design could be used across platforms handling central bank money, stablecoins and real-world assets. That is the basis for reading it across. It is a prototype, and the report says plainly that it is not production-ready.

Privacy enforced, not promised

The report's organising principle is a single sentence: privacy should be technologically enforced rather than institutionally promised. In other words, a user's protection should not depend on a central bank or an intermediary saying it will not look, but on a system in which it cannot.

The design separates who knows what. The bank that onboards a customer performs know-your-customer checks and holds the customer's legal identity. The central bank cannot access the retail ledger and therefore cannot see who anyone is. Each bank's ledger is also partitioned from every other bank's, so no single party can reconstruct a payment history. Payments are validated with zero-knowledge proofs, which hide amounts and the link between payer and payee from parties that do not need them. Users transact under evolving pseudonyms that are unlinkable to their real identity but verifiable as belonging to a KYC-checked account, so a merchant cannot profile spending even by merging ledgers with other merchants.

Compliance is retained, but at the edge. When a payment meets or exceeds a bank-defined threshold, the payer's identity is disclosed automatically, in masked form. Law enforcement can resolve any pseudonym to a real person by going to the issuing bank, an approach the report links to FATF's provision on relying on third parties for customer due diligence.

The inverse of a public blockchain

Set that design next to how stablecoins circulate today, and it looks like a mirror image.

A public blockchain hides who you are and shows everything else. Addresses are pseudonymous, but every amount, every transfer and every counterparty address is visible to anyone, permanently. The report lists public blockchain tracking among the privacy concerns its design addresses. Aurum does the reverse: identity is held firmly by a regulated bank, while amounts and transaction links are hidden from everyone who does not need them.

That inversion matters for supervision as well as for users. When we looked at the FSB's recent note, the problem was that regulators want to know where stablecoin holders are, and the FSB's own data review says that is often not possible on public chains. Aurum's answer is to stop trying to infer identity from the ledger and anchor it at the intermediary instead, where it can be produced on lawful demand. The public chain gives everyone partial information. Aurum gives the right party complete information and everyone else almost none.

The controls are finer-grained than an on-off switch

Three features stand out because they have direct analogues in today's stablecoin market.

Fractional locking. Banks today can freeze an account or put a hold on part of a pooled balance. Aurum can lock an individual coin, a subset of coins, or a whole wallet. If one incoming payment looks suspicious, only the coins from that transfer are locked and the rest of the wallet stays usable. On public chains, stablecoin issuer freezes typically apply to an address as a whole. Coin-level locking is a more proportionate tool.

Anonymous wallets, with limits. The report accepts that full anonymity sits uneasily with compliance, and treats it as a calibrated exception. Its example is tourists, who may want minimal payment trails and be reluctant to hand over passport data. An anonymous wallet with low balance and spending limits, bound to a device and issued by a single neutral intermediary such as a post office, gives cash-like privacy for small payments while preventing one person from opening many.

Configurable onboarding. Rather than a single KYC standard, the prototype offers a menu: no identity checks with tight caps, device binding, one-time eligibility proofs, a third-party identity service such as Hong Kong's iAM Smart, proofs from a phone's secure enclave, or full checks with stored data. Each jurisdiction chooses its own point on the curve.

Every privacy gain is paid for somewhere

The report is unusually direct about costs, and they are worth stating because they apply to any token that tries to borrow this model.

Speed. Zero-knowledge proofs are quick to verify but slow to generate, and proof generation is the main bottleneck. GPU acceleration helps. An optional batch mode, useful for merchants consolidating many small payments, improves efficiency but lets the validator link the coins in a batch, so it buys speed with privacy.

Programmability. Because counterparty information is not visible in real time, Aurum cannot offer general-purpose smart contracts. Programmability is limited to conditions embedded in individual coins and logic enforced by banks off-chain. Some conditional payments require the user to delegate temporary custody to a bank, which reintroduces counterparty risk. The report cites Brazil's Drex experiments as reaching the same conclusion: stronger privacy impaired programmability.

Recovery. At the highest privacy setting, nothing links a wallet to its owner, so lost funds cannot be recovered. The report compares this directly to losing physical cash or losing the private key to a crypto wallet.

Portability. A user's access is tied to the bank that onboarded them. If that bank becomes unavailable, the user may lose access even though their pseudonymous presence remains valid. The report suggests verifiable credentials as a way to make access portable between banks, but leaves it as future work.

None of these are reasons to abandon the approach. They are the choices a jurisdiction has to make explicitly rather than inherit by default, which is the same framing the Bank of Italy's financial intelligence unit used when it proposed constrained privacy for stablecoins. Two very different institutions have now arrived at the same broad shape of answer: privacy as the default for ordinary payments, bounded by compliance controls rather than abandoned to them.

What follows for each group

Stablecoin issuers. Issuers that already sit inside a regulated banking group hold the piece Aurum's design depends on most: customer identity held by a licensed institution. Whether a privacy layer of this kind could be applied to a reserve-backed token running on a public chain is untested, and the report does not claim it can. But the design shows what regulators may come to see as achievable.

Banks. In this model the bank is where identity, compliance and lawful access live, which makes the intermediary more central rather than less. The portability problem is the flip side: a user tied to one bank's onboarding is tied to that bank's availability.

Regulators. The prototype treats privacy level as a set of dials (disclosure thresholds, wallet limits, onboarding tiers, locking granularity) rather than a binary. That gives supervisors a vocabulary for proportionate rules that today's stablecoin regimes, which mostly address issuance and reserves, have not yet developed. Tracking how each jurisdiction is drawing those lines is part of why we maintain the free Stablecoin Regulation Tracker across 200+ markets.

Policymakers. The e-HKD consultation found privacy to be the public's biggest concern, ahead of cybersecurity, performance and resilience, with most respondents saying it should take precedence over efficiency. The trade-offs the prototype documents make that preference expensive in specific, measurable ways.

The question carried forward

Hong Kong has always been comfortable with banks issuing its money, provided that money is fully backed and tightly supervised. Its first licensed stablecoins extend that tradition, though they are backed by high-quality reserve assets under the Stablecoins Ordinance rather than by central bank money as Phase 1 imagined.

What the tradition has never had to settle is privacy, because paper notes provide it automatically. A banknote does not report where it was spent. A token on a public blockchain reports everything except the spender's name. Aurum Phase 2 shows a third option exists, at a cost in speed, programmability and recoverability. The open question for bank-issued digital money is no longer whether that option is technically possible. It is which of those costs each jurisdiction is willing to pay.

Source: Project Aurum Phase 2: Privacy-enhancing technologies for digital cash, BIS Innovation Hub Hong Kong Centre, with the Hong Kong Monetary Authority, the Hong Kong Polytechnic University and the Hong Kong Applied Science and Technology Research Institute, 2026. The report describes its prototype as exploratory and not production-ready.

Additional data: Phase 1 design details, including the description of the CBDC-backed stablecoin as the digital equivalent of the note-issuing bank system, from Project Aurum: a prototype for two-tier central bank digital currency, BIS, 21 October 2022. Note-issuing banks and the HK$7.80 link to US dollars from the HKMA's Hong Kong currency pages. Stablecoin issuer licences granted to HSBC and Anchorpoint Financial from the HKMA press release of 10 April 2026; Anchorpoint's Standard Chartered backing as reported in legal commentary on the licences. The description of licensed stablecoin backing follows the HKMA's own wording in its 2026 Project Ensemble business use case report.

Methodology: We focused on the prototype's privacy architecture, its controls and its stated trade-offs. Phase 2 tests a central bank liability token; applying its design to bank-issued stablecoins is our extension, supported by the report's single reference to stablecoins in its discussion of cross-platform pseudonyms. The comparison with public blockchain stablecoins is our own.

This article is for informational purposes only and does not constitute financial, investment, or legal advice.