Client Asset Protection Map: Deposits, Safeguarding and Investor Compensation

Concept

Three different bodies of law answer the question "what happens to my money if the provider fails", and they answer it with three different instruments. A bank deposit is covered by a guarantee scheme that pays a fixed sum out of a fund the industry has already contributed to. An e-money or payment balance is covered by nothing of the kind: what exists is a segregated pool and a rule that puts the customer first in the queue against it. A brokerage position is covered mostly by the fact that the asset never entered the intermediary's balance sheet, with a compensation scheme sitting behind that as a small backstop for property that has gone missing.

The three are routinely spoken of as though they were versions of one idea. They are not, and the difference decides which provider is appropriate for which money. The table below sets the regimes on one axis.

Read down the "who pays" column and the practical hierarchy appears. Only the first row involves somebody else's money arriving to make the customer whole. The second and third rows describe rules about the customer's own property — where it sits, whose estate it is in, and who gets to it first. That is why a bank with a modest limit can be safer for cash than a payment institution with an impeccable balance sheet, and why a broker with no compensation scheme at all can be safer for securities than one with a generous limit.

The mechanics of the payment chain and of the safeguarding methods themselves belong to correspondent banking and client-money protection; what happens as a licence is surrendered belongs to licence withdrawal and wind-down. This page is the map: the schemes, the limits, the exclusions and the allocation decision that follows from them.

Deposit guarantee schemes on one table

The limit is per depositor per licensed entity unless the row says otherwise, and the last column gives the date of the primary source behind the row. Two axes are usually missing from the numbers as they are quoted: which currencies the scheme actually covers, and who it recognises as a depositor at all. Both change the answer more often than the headline figure does.

Three patterns run through the table and none of them is about the size of the limit. The first is currency: Singapore covers Singapore dollars and nothing else, Türkiye states its limit in lira, Bermuda covers Bermuda dollars, and Kazakhstan splits the ceiling by currency and puts the foreign-currency tranche at a quarter of the tenge one. A multi-currency account at a scheme member can therefore be entirely uninsured in the currency the client actually holds. The second is the eligible depositor: Jersey and Guernsey exclude ordinary companies and trusts outright, the Isle of Man admits them at a fifth of the individual limit, Kazakhstan excludes legal entities altogether. A structure — a holding company, a discretionary trust, a private trust company — frequently has no cover at any limit, which is the point made at length in offshore banking jurisdictions. The third is the absence of a scheme: the UAE at federal level, Puerto Rico's international banking entities, Dominica and Panama have none, and the DIFC and the Cayman Islands substitute a place in the insolvency queue for a fund.

Payout speed is the axis that is quoted most confidently and delivered least uniformly, because "payment" can mean a transfer to a liquidator, the dispatch of an instrument, or money the customer can spend.

The remaining schemes in the map publish no target, and the per-scheme qualifications, the branch-and-subsidiary question and the temporary-high-balance rules are worked through in personal bank accounts abroad. What the deposit layer does not answer at all is what a bank's own capital and the insolvency queue do above the limit, which is the subject of bank reliability.

Safeguarding: what a payment licence secures and what it does not

Nothing in the payment world resembles a guarantee scheme. Safeguarding is a duty imposed on the operator to keep customer money somewhere other than its own working capital, and the customer's protection is the quality of that arrangement plus a place in a queue. The regimes differ in what the pool consists of, whether a trust sits over it, and — decisively — who is obliged to make up a shortfall. Nobody is.

Two failures define what the layer actually delivers, and they fail in opposite directions. The first is the ledger failure. When Synapse Financial Technologies, a middleware provider through which several US neobanks reached partner banks, collapsed in 2024, roughly US$219m of end-user money was frozen in May of that year and the reconciliation of who owned what inside pooled accounts rested on Synapse's own books. Those books diverged from the banks' balances by somewhere between US$65m and US$95m; more than a hundred thousand people were affected, and the partner bank declined to close the gap out of its own funds. The funds were in the right kind of account and the segregation was formally in place — what was missing was a reliable record of whose money it was. The FDIC's recordkeeping response to that episode remained a proposed rule as at 17 August 2026. The structure that produced it is analysed in BaaS: how a banking product lives without a licence.

The second is the freeze. When the FCA imposed requirements on Wirecard Card Solutions Ltd on 26 June 2020 — the firm was not to dispose of any assets or funds and was not to carry on regulated activities — the cards of a long list of consumer apps stopped working over a weekend. The regulator consented on 29 June and the restriction lifted at 00:01 on 30 June 2020, four days later, once the firm's compliance with the safeguarding rules had been verified. Nobody lost money to a shortfall. What was demonstrated was the other exposure: an intact pool is of no use to somebody who needs to pay for groceries on Saturday, and a payment balance is not a store of value even when it is safeguarded correctly.

The arithmetic of a genuine shortfall, the recoupment of funds that ought to have been safeguarded and the pence-in-the-pound outcome are worked through by the owner of the mechanics at correspondent banking and client-money protection.

Investor compensation: the smallest number in the chain

The securities layer works the other way round from the deposit layer. The primary protection is that the asset is not the intermediary's: client securities are segregated and, on a failure, are returned past the estate. A compensation scheme sits behind that as a backstop for property that has gone missing — not as insurance on the portfolio. That is why the limits look absurd next to real portfolios and why it is a mistake to choose a broker on them.

The column that decides anything is the fourth one. Every scheme in the table excludes a fall in market value and bad advice; several exclude the professional client, the firm's own insiders and anybody holding 5% or more of it; and Canada, the Netherlands and the United States each exclude crypto-assets in terms. The limits themselves span two orders of magnitude — €20,000 in Ireland, Luxembourg, Cyprus and the Netherlands against C$1,000,000 in Canada — and on a seven-figure portfolio all of them are rounding errors. What is not a rounding error is whether the securities leave the intermediary's balance sheet: Swiss law issues them to the client regardless of the bank's fate, MiFID II forbids an investment firm from taking title-transfer collateral from a retail client, and a US margin account expressly permits the broker to reuse pledged stock. The entity-by-entity version of that analysis, with the access and tax consequences of each booking, is in international brokerage accounts; the custody chain behind it is in securities custody.

Crypto: why there is no segregation layer to speak of

The crypto regimes have imported the language of segregation without importing anything behind it. Under Regulation (EU) 2023/1114, a crypto-asset service provider holding clients' crypto-assets must keep them unencumbered, must not use them for its own account, must make adequate arrangements to safeguard clients' ownership rights and must maintain a custody policy available to clients on request. Those are conduct obligations on the provider. No compensation fund stands behind them, no guarantee scheme responds to the provider's failure, and the regulation creates neither.

What makes the gap concrete is that the conventional schemes have written the exclusion down. The Dutch investor-compensation scheme states that crypto-assets are not protected under it. The Canadian fund does not cover crypto assets held by a member firm that are missing at the time of the firm's insolvency — the exact circumstance a compensation scheme exists for. SIPC does not protect a digital or crypto asset that does not qualify as a registered security, and the statutory definition it works from excludes stablecoins and currency outright. Three regulators in three legal systems have been asked the question and have answered it the same way.

The practical consequence is that a crypto position sits in the second column of the opening table and not the first or third: the protection is the quality of the custody arrangement, the identity of the custodian and whether the assets are genuinely segregated both on-chain and in the books, with nothing behind it if they are not. The custody structures that answer that question, and the difference between an exchange balance and a qualified custodian, belong to crypto for private wealth; the licensing perimeter is in the MiCA CASP licence.

Where to keep which money

The allocation follows from the three regimes rather than from a ranking of providers. Each row below answers one question: what is this money for, and which regime is appropriate to that purpose.

A worked example on US$2,000,000

Assume an individual, resident outside the EU and the United States, holding US$2,000,000 across the providers a globally mobile client typically accumulates. The conversions use round illustrative rates of £1 = US$1.35 and S$1 = US$0.78, and are arithmetic rather than market data.

Of the US$850,000 held as cash, US$412,000 is insured — 48% of the cash and 21% of the total. Three moves change that without changing a single asset class. Splitting the UK holding across two unconnected banking authorisations takes the covered amount there from US$162,000 to US$324,000. Converting the Singapore balance into Singapore dollars brings S$100,000, about US$78,000, inside the scheme, and moving the surplus to a second US bank in a different ownership category covers it in full. Moving all but a working float off the e-money balance removes US$150,000 from a pool with nobody behind it. Insured cash rises from US$412,000 to roughly US$652,000, and the sum exposed to an unguaranteed pool falls from US$150,000 to a month's spending. The US$900,000 of securities was never the problem: its protection is the segregation regime, which is unlimited in amount and worth more than every compensation limit in the map put together.

What goes wrong most often

• Counting protection by brand. Cover attaches to a licensed legal entity. One group's UK, Irish, Swiss and Singapore entities carry four different schemes, four limits and four insolvency laws; the app looks identical in all four cases.
• Reading a safeguarding label as insurance. "Funds are safeguarded" and "held at an insured bank" describe where the money sits, not who pays if the operator fails. The answer there is: nobody, beyond what the pool contains.
• Holding the wrong currency at the right bank. Singapore covers Singapore dollars, Türkiye lira, Bermuda Bermuda dollars, and Kazakhstan splits the ceiling and leaves foreign currency at a quarter of the tenge limit.
• Choosing a broker on its compensation limit. €20,000 against a seven-figure portfolio decides nothing. Segregation, rehypothecation terms and the custody chain decide everything.
• Treating a temporary high balance as protected. The UK's £1.4m applies for six months after a qualifying event; it is a bridge after a house sale or an inheritance, not a standing limit.
• Assuming a structure is a depositor. A company, a trust or a private trust company is outside the eligible class in several schemes at any limit, and moving the money into a structure can remove cover the individual had.
• Expecting a scheme to answer a freeze. No guarantee scheme, safeguarding pool or compensation fund responds to a sanctions measure, a regulatory restriction on the operator or a correspondent's exit. Those are availability events, and they are more common than insolvency.