Four jurisdictions, four philosophies. The GENIUS Act formalized US payment stablecoins in November 2025. MiCA reached full operative status across the EU in 2024. Singapore MAS expanded Project Guardian to include retail pilots in 2025. The UK Digital Securities Sandbox opened the same year. Headlines called all four "stablecoin frameworks" or "tokenization regimes." The actual texts ask for different things.
The thread that follows compares the four on four axes: reserve composition, transfer agent and custody neutrality, bankruptcy remoteness, and asset eligibility. The pattern is clear: each jurisdiction is the easiest place to ship a specific kind of product. Where you incorporate the issuer is a function of what you are trying to issue.
GENIUS Act, the US payment-stablecoin regime
Federal preemption, narrow eligible reserves, bankruptcy remoteness via property segregation. The most prescriptive of the four, scoped tightly to one product.
The GENIUS Act, signed November 2025, did one thing forcefully: it preempted state money-transmission regimes for payment stablecoins and replaced them with a federal regime administered through the OCC and a new "permitted payment stablecoin issuer" designation.
Eligible reserves under the Act are exhaustively listed: cash held at insured depositories, short-term US Treasuries, repo collateralized by Treasuries, money market fund shares whose own holdings meet the same criteria, and tokenized equivalents of all of the foregoing that meet specified custody and segregation requirements. The list excludes commercial paper, long-duration bonds, non-USD assets, and any equity exposure. The Act explicitly forbids using stablecoin reserves to make unsecured loans.
Transfer agent neutrality is not required. A licensed issuer can maintain the holder register internally (the way Circle does for USDC) or appoint an external transfer agent (the way BlackRock appoints BNY Mellon for BSTBL). The Act is silent on which architecture is preferred.
Bankruptcy remoteness is achieved through a specific property-law clause: reserves are deemed the property of token holders rather than the issuer, so a bankruptcy of the issuer cannot reach the reserve assets. The mechanism is statutory rather than contractual, which gives it more protection than private agreements would.
The cost of this regime is asset eligibility. A GENIUS-compliant stablecoin cannot include private credit, commodities, real estate, or equities in its reserves. It is explicitly a payment instrument backed by sovereign debt and nothing else. That makes the US the easiest place to ship a Treasury-backed payment stablecoin at scale, and the hardest place to ship anything more exotic.
MiCA, the EU service-provider regime
Three product categories, three licenses, MiFID-equivalent investor protection. The most comprehensive of the four, scoped broadly across the whole crypto stack.
MiCA (the EU Markets in Crypto-Assets Regulation, fully operative 2024) is structurally different from GENIUS in that it regulates the service providers and the products in one framework rather than treating them separately.
Transfer agent neutrality is required for ART issuers but not for EMTs. The asymmetry is intentional: ART holders need independent verification of the basket composition, while EMT holders are protected by the 1:1 face-value redemption promise.
Reserve composition rules are stricter than GENIUS for ARTs: the European Banking Authority specifies high-quality liquid asset categories with concentration limits per asset, per issuer, and per maturity. EMTs follow electronic-money directive rules that are looser than the GENIUS list. The practical effect is that MiCA makes the EU the easiest place to issue a non-USD stablecoin or a multi-asset stablecoin, and a slightly harder place than the US to issue a pure USD-Treasury stablecoin.
Bankruptcy remoteness is achieved through segregation rather than property assignment. Reserves must be held in separate accounts at credit institutions, and the regulator can require additional capital cushion. The protection is strong but more administrative than the GENIUS statutory clause.
Singapore MAS, the sandbox-by-product regime
Project Guardian approves pilots individually with bespoke rules. The most flexible of the four, designed for products the other regulators have not seen yet.
MAS Project Guardian runs differently from the rule-based frameworks above. It is a sandbox: financial institutions propose specific pilots (tokenized assets, tokenized funds, tokenized FX, programmable money) and MAS approves each one with bespoke conditions.
By 2025, Project Guardian had expanded to phase 3 with retail-eligible pilots, multiple chain deployments, and over twenty participating financial institutions including JPMorgan, Apollo, UBS, Standard Chartered, and DBS. The framework is explicitly forward-looking: products are approved that would not yet have a clear home under GENIUS or MiCA.
Reserve composition rules vary per pilot. Some tokenized money market fund pilots use the same composition as a BUIDL-equivalent. Others, particularly the FX pilots, hold non-USD reserves that GENIUS would not allow and that MiCA would categorize as ARTs requiring a separate license.
Transfer agent neutrality is encouraged but not mandated. Many pilots use approved transfer agents from a MAS-curated list, but the framework permits issuer-maintained registers with additional disclosure.
Bankruptcy remoteness follows existing Singapore trust law. Reserves are held by a licensed trustee for the benefit of holders. The mechanism is mature and well-tested in the existing fund-management industry, which is one reason institutional issuers find Singapore comfortable.
The cost of Project Guardian’s flexibility is unpredictability: every new product requires a separate approval, and a pilot’s terms cannot be extrapolated to other applicants. The framework is excellent for first-of-kind issuance and poor for scaling commodity products. Most participants use Singapore to prove out a novel structure and then re-issue under MiCA or GENIUS when they want scale.
US is easiest for payment stablecoins. EU is easiest for multi-asset and non-USD stablecoins. Singapore is easiest for first-of-kind. UK is easiest for tokenized public securities. The arbitrage is structural, not temporary.
UK Digital Securities Sandbox, the opt-in securities regime
Joint FCA-BoE framework that lets approved issuers ship tokenized securities under bespoke rules for up to five years. Targeted narrowly at the asset class GENIUS does not cover.
The UK Digital Securities Sandbox (DSS), operative since 2024, is a joint Financial Conduct Authority and Bank of England framework. It allows approved firms to issue, trade, and settle tokenized securities (equities, bonds, fund units) under modified rules for up to five years, after which the firm migrates to the permanent regime that the sandbox produces.
The DSS is structurally narrower than GENIUS, MiCA, or Project Guardian. It does not regulate stablecoins (those run under the UK’s separate stablecoin framework via the FCA). It does not regulate utility tokens. It regulates only tokenized securities, which is precisely the asset class that GENIUS does not address and that MiCA addresses only indirectly through MiFID.
Reserve composition rules are not applicable in the traditional sense (securities are not reserved against a fiat unit). Instead, the DSS requires issuers to maintain a registry of ownership that is functionally equivalent to a Central Securities Depository record, and to clear settlement through an approved post-trade infrastructure.
Transfer agent neutrality is allowed in a way that distinguishes the UK regime: the issuer can maintain the ownership register directly through the blockchain itself, with no separate transfer agent, provided the chain meets specified operational resilience standards. That gives onchain-native issuance a regulatory home in the UK that does not exist cleanly anywhere else.
Bankruptcy remoteness follows existing UK securities and custody law, which has bankruptcy-remote nominee structures that the industry has used for decades.
The arbitrage map
Where each rule is strictest, where it is loosest, and which products belong where.
Stack the four regimes by axis and the arbitrage becomes legible.
What every regime still lacks
Three things none of the four addresses cleanly.
Across the four regimes, three gaps remain that constrain what tokenization can do.
First, none of the four addresses cross-jurisdictional asset recognition cleanly. A token issued under MiCA as an EMT is not automatically valid as a GENIUS stablecoin for US regulatory purposes, and vice versa. Cross-border issuers must maintain parallel structures (a USDC equivalent on each side of the Atlantic, effectively) until mutual-recognition treaties catch up. The industry has lobbied for this for years; the diplomatic timeline runs in decades.
Second, none of the four has a settled rule for tokenized private assets at retail scale. Private credit, private equity, and real estate tokenization remain accredited-investor-only in the US, MiFID-professional in the EU, sandbox-pilot in Singapore, and not yet addressed by the UK DSS for tokenized private securities. The retail unlock for these asset classes is the next regulatory frontier; none of the four has cleared it.
Third, all four regimes assume the chain itself is a neutral infrastructure. None of the four explicitly licenses the blockchain operators, validates the consensus mechanism, or prescribes operational resilience standards on the underlying L1 or L2. The UK DSS comes closest, but only for the issuer’s operational resilience, not the chain’s. When the next major chain outage hits a tokenized fund, every regime will scramble to figure out what to require.
Until those three gaps close, the regulator scoreboard reads as it does today: each jurisdiction wins a specific product category, and a competent issuer routes its different products through different jurisdictions to land each one in its easiest regime. The map will keep showing the asymmetry until the rules converge, which on current evidence will not be soon.