Singapore stablecoin rules for merchants: MAS defined the label on September 1, not the checkout

What MAS actually published on September 1, 2026

Singapore has had a stablecoin framework on paper since August 2023, when MAS published its response to the 2022 consultation and set out the MAS-SCS regime for single-currency stablecoins. What it did not have was statute. On September 1, 2026, that changed: MAS opened a consultation on draft legislative amendments to the Payment Services Act 2019 that turn the framework into enforceable law, plus a set of new proposals responding to how the market moved between 2023 and now.

The core architecture is straightforward. Issuance of a single-currency stablecoin becomes a regulated activity with its own licence class, sitting beside the existing money-changing, standard payment institution and major payment institution licences. Issuance is defined broadly enough to catch the incidental work — minting, putting tokens into circulation, managing reserve assets, redeeming at par — so an issuer cannot unbundle its way out of the perimeter.

The scope is deliberately narrow. The regime covers stablecoins issued in Singapore and pegged to the Singapore dollar or any G10 currency. Multi-currency and basket tokens are out. Everything that does not qualify remains perfectly lawful — it simply keeps being treated as a digital payment token (DPT), with intermediation regulated under the existing DPT licensing regime rather than the stablecoin one.

Then the substantive bar. A licensed issuer would have to hold reserve assets at least equal to the par value of tokens in circulation at all times, in liquid low-risk instruments, segregated from the issuer's own balance sheet, held with approved custodians and subject to independent attestation. Redemption at par is a statutory obligation, with the operational window set out in subsidiary legislation to come — the 2023 framework's five-business-day commitment is the reference point the market has been working to. Add capital and solvency standards, disclosure obligations, safeguarding of customer monies received before tokens are issued, quarterly stress testing, and board-approved recovery and orderly wind-down plans reviewed annually.

Finally, three provisions that did not exist in the 2023 policy and that are the real news: a flat prohibition on paying interest, a designation power over systemic stablecoins that reaches foreign tokens, and a recognition regime for foreign stablecoins governed by substantively equivalent frameworks. Those are the ones worth a merchant's time.

Read it as a business that gets paid, not as an issuer

Here is the exercise. Take the draft and ask the four questions a business actually has about a payment instrument: who accepts it, what does accepting it cost me, can the payment be reversed, and where does the money land?

The draft answers none of them. It is not a flaw in the drafting — the Payment Services Act amendments are an issuance rulebook, and issuance is what they regulate well. But that is now the fifth consecutive major jurisdiction to publish a serious stablecoin regime addressed entirely to the supply side. The US GENIUS Act, the EU's MiCA, Japan, Hong Kong, South Korea and now Singapore have collectively written thousands of pages about who may print a digital dollar and under what conditions. The number of pages devoted to how a shop, a clinic, a studio or an API company actually gets paid in one is approximately zero.

The market data says the same thing from the other end. Research by Flagship Advisory Partners put direct stablecoin acceptance at roughly 4% of top-50 US merchants, 8% in Europe and 12% in Latin America, against retail stablecoin volume of around $70 billion in 2025 — real growth, but next to more than $50 trillion a year across the major card networks. Issuance is crowded and getting more regulated by the month. Acceptance is still mostly empty.

So what a Singapore merchant gets out of September 1 is a label to check — useful, genuinely — and no new way to take a payment. That gap is the whole subject of the rest of this piece.

Three provisions that land on merchants even though none is addressed to them

1. "Trace, freeze and burn" is the phrase to read twice

The draft's AML and CFT expectations include the ability to trace, freeze and burn tokens. This is not novel — every major fiat-backed issuer already has an administrative freeze function, and it has been used. What is new is that Singapore proposes to make it a condition of the licence, which turns a discretionary capability into a supervised obligation.

For a business that holds a balance, the relevant question shifts. It stops being "is this token compliant?" and becomes "how many parties can touch my money after the sale is done?" Those are different questions with different answers.

Be precise about what a payment processor can and cannot do here. An issuer's freeze power is a property of the token itself; no processor removes it, and anyone who tells you otherwise is selling something. What a non-custodial processor removes is a different layer: the intermediary balance. When funds go straight from the payer to a wallet the merchant controls, there is no processor float, no pooled omnibus account, and no scenario where a business's money is frozen because it happened to be sitting in the same pot as somebody else's. In custodial crypto processing, that pot is the single largest operational risk a merchant carries, and it has nothing to do with regulation — it is just where the money was standing when something went wrong.

The second mitigation is not being single-issuer dependent. Accepting more than one token, on more than one chain, with optional auto-conversion to USDC or USDT, means an issuer-level problem is a routing decision rather than a treasury event.

2. The interest ban settles the float argument permanently

MAS proposes to prohibit issuers from paying interest, returns or any benefit attributable — directly or indirectly — to holding an MAS-regulated stablecoin. The stated rationale is clean: these instruments are meant to be payment and settlement tools, not deposits or investments. There is a carve-out for commercial arrangements where a third party pays rewards to its own customers out of its own account, which is a distribution incentive for exchanges and wallets, not revenue for the business accepting the token.

This closes an argument that has wasted a lot of merchant attention. Every jurisdiction that has written a serious framework — the US, the EU, and now Singapore — has landed in the same place: the yield on the reserves belongs to the issuer. That is the deal. A business should therefore stop evaluating stablecoins on a number that will never be theirs, and evaluate them on the three that are:

• Cost. Network fees measured in cents on Solana, Base or Polygon, instead of a percentage of every sale plus cross-border and FX surcharges.
• Speed. Confirmation in roughly 0.4 seconds on Solana and about 2 seconds on Base and Polygon, instead of a one-to-three-day settlement cycle with a weekend in it.
• Finality. A confirmed on-chain payment is not reversible by the payer, which is the structural difference from a card transaction that can be pulled back for roughly 120 days.

Those three are worth far more to an operating business than any share of reserve yield would have been, and unlike the yield, nobody is proposing to legislate them away.

3. Designation reaches tokens issued outside Singapore

The proposed Part 2A is the sharpest instrument in the draft. It would let MAS designate a stablecoin as a Designated Systemic Stablecoin whether it is issued inside or outside Singapore, and whether or not it is MAS-regulated at all. Designation factors are the familiar ones: size in circulation, interconnectedness with payment systems and the wider financial system, and substitutability. Where a designated token fails to meet requirements, MAS could direct licensed DPT service providers in Singapore to stop offering it, delist it and its trading pairs, and prohibit customers from accumulating more of it.

Read that against what happened in Europe eight days before this draft was published, when a major consumer fintech delisted USDT for its EU users and required balances to be converted. Same pattern, different legal route: the token you accept can lose its local venue without your business doing anything wrong. Your customers' ability to buy it changes, your off-ramp options change, and if your money was sitting on that venue, so does your access to it.

There is a genuinely constructive counterpart in the draft. MAS also proposes a recognition regime for a limited number of foreign stablecoins governed by substantively equivalent frameworks with supervisory cooperation arrangements, plus a mechanism for jointly issued tokens where affiliated entities in different jurisdictions issue against common reserves. That is the first real outline of something like passporting for stablecoins, and it is the right long-term direction. It is also case-by-case, discretionary, and years from mattering to a shop in Singapore. Between now and then, fragmentation keeps growing.

The merchant-side answer: acceptance that doesn't bet on an issuer

If the pattern across every framework is that regulation attaches to issuers and acceptance is left to the market, then the design brief for a business is obvious. Do not build your payments around a single token, a single chain, or a single licensee's continued good standing in your jurisdiction. Build around an acceptance layer that treats tokens as interchangeable and sends the money somewhere no third party can reach.

That is what Payzum is. It is non-custodial and crypto-only: funds go directly to wallets the merchant controls, and Payzum never holds, pools or controls the money. Settlement is the payment — there is no payout queue, no reserve, no rolling hold, and no Payzum balance for anyone to freeze. Auto-conversion to USDC or USDT is optional, so a business can accept whatever a customer holds and still keep its books in dollars.

The concrete capabilities, all of which exist today:

• Online: no-code payment links and buttons, hosted checkout (redirect, modal or inline), invoices with expiry and overpayment detection, donation and tip-jar flows, and recurring subscriptions. Drop-in compatible with existing e-commerce plugins, snippets and webhooks.
• In person: a POS that turns any phone into a terminal — a fresh QR per sale, physical terminals where you want them, PIN-protected cashier accounts with per-cashier and per-terminal analytics. No acquirer, no card-network fees, no chargebacks.
• Payouts: mass payouts by CSV (BTC, LTC, DOGE) and EVM stablecoin payouts across Polygon, Arbitrum, Optimism, Base, BNB Chain and Avalanche.
• Developers and AI agents: a REST API with API keys, signed webhooks, an integration playground, and x402 so AI agents can pay USDC on Base per API call, straight to the merchant's wallet.
• Networks: Bitcoin, Ethereum, Solana, Polygon, Base, Arbitrum, Optimism, BNB Chain and Avalanche, plus LTC and DOGE for payouts.
• Security: 2FA, signed webhooks, encrypted secrets and a full audit log.

On x402, one point of precision because it is widely misreported: Payzum is the middleware and proxy in front of your API, not the facilitator. You configure your existing endpoint, your API key or bearer token, and a price. Payzum publishes an x402 URL, returns the 402, settles the payment through an external facilitator (currently Coinbase's), then proxies the paid call to your real endpoint with your key. No protocol to implement and no code to write — a dashboard configuration, and an API provider can start serving paying agents the same day.

How you'd actually set this up, step by step

1. Connect a wallet you control. Your own custody, your own keys. Choose your settlement chains — Solana, Base and Polygon are the usual picks for retail-sized payments because confirmation is sub-second to about two seconds and fees are cents. Turn on auto-conversion to USDC or USDT if you want every sale to land as dollars.
2. Pick the surfaces you need. An online store gets a hosted checkout or a drop-in plugin; a services business gets payment links and invoices with expiry; a physical location gets the POS app with a fresh QR per sale; a subscription business gets recurring billing. Most businesses turn on two or three.
3. Set up your people and your systems. Create PIN-protected cashier accounts so staff can take payments without touching wallet settings, and wire signed webhooks into your ERP, accounting or fulfilment system so a confirmed payment triggers whatever it should trigger. Every action lands in the audit log.
4. Go live and watch settlement, not payout. There is no payout to wait for. The customer's confirmed transaction is the settlement, arriving in your wallet in seconds. Reconciliation runs per cashier, per terminal and per checkout.

What this looks like for four kinds of business

Singapore is a useful test case because it is small, open, and its merchants trade across a dozen currencies before lunch. These are the patterns we see most often.

• A Singapore e-commerce brand selling into Southeast Asia. Card acceptance across Indonesia, Vietnam, the Philippines and Thailand means multiple acquirers, uneven approval rates, FX on every leg and cross-border interchange. A hosted checkout that takes USDC or USDT on Solana or Base collapses that into one flow with one settlement currency, arriving in seconds regardless of which country the buyer is in — and no chargeback window trailing behind each order.
• A retailer or restaurant taking tourist money at the counter. A visitor holding stablecoins pays by scanning a QR generated fresh for that sale; the cashier is on a PIN-protected account on an ordinary phone; the money is in the owner's wallet before the receipt prints. No terminal rental, no acquirer relationship, no dynamic-currency-conversion games, and nothing to charge back later.
• A Singapore API or AI company selling to machines. An agent that wants your endpoint gets a 402, pays USDC on Base for that single call, and gets the response — with the payment landing in your wallet. You configured an endpoint and a price; you did not implement a protocol. This is the segment where "who accepts it" is being decided right now, and it is decided by whoever is configured to take the payment.
• A regional agency or studio paying contractors across the region. One CSV, one batch, contractors in six countries paid in stablecoins on Polygon or Base for cents in fees, without a correspondent bank taking three days and a cut on each leg.

Custodial crypto processing vs non-custodial acceptance, under this kind of regime

Two objections worth answering directly

"Shouldn't I just wait until the regime is in force and only accept MAS-regulated tokens?"

Two problems with waiting. First, timing: the consultation closes October 16, 2026, subsidiary legislation is a separate consultation still to come, and licensing then has to happen — this is a multi-year path, and businesses that wait will spend those years paying full card economics. Second, and more important, the label answers a question you were not asking. "MAS-regulated" tells you about the issuer's reserves and governance. It does not tell you who will accept the token at your checkout, what acceptance costs you, whether the payment can be reversed, or where the money lands. Those remain merchant-side decisions no matter which tokens end up carrying the label. Non-custodial, token-agnostic acceptance is compatible with every outcome of this consultation, which is precisely the point of building it that way.

"If issuers can freeze and burn, isn't self-custody a false comfort?"

No, but it is a narrower comfort than it is often sold as, so let's be exact. Issuer freeze powers are real and they attach to the token; self-custody does not override them, and no processor can. What self-custody removes is the far more common failure mode: your funds sitting in someone else's account when that account is frozen, disputed, delayed or wound down. Look at the actual history of merchants losing access to crypto revenue and it is overwhelmingly intermediary failure, not issuer action against an ordinary business. The layered answer is straightforward — hold your own keys, accept more than one token across more than one chain, and use auto-conversion so your working balance is in the instrument you chose rather than the one a customer happened to pay with.