Abstract

This article provides a comprehensive dataset on the terms and conditions (T&Cs) set by leading providers of crypto custody and wallet services. The sample studied reveals a dichotomy within the crypto custody industry: some custodians seek to protect their clients’ rights and interests while others provide T&Cs at odds with any meaningful client protection. In such an environment, financial regulation can serve three purposes. It can promote a level playing field for crypto custodians, ensure an adequate level of client protection regardless of client due diligence, and reduce the likelihood of costly disputes, thereby lowering both information asymmetry and transaction costs.

This article provides an overview of how crypto-assets are held in custody before outlining the research sample. The article then presents findings on custody practices, including safekeeping, key storage, outsourcing, asset segregation, insolvency protection, client entitlements, reuse, liability cap, and choice of law, courts, and arbitration. Finally, the argument and conclusion are set out.

I. INTRODUCTION

Custodians are a crucial cornerstone of the USD 500 billion digital asset market.1 This article analyses the terms and conditions (T&Cs) of providers of crypto custody and wallet services. It provides an overview of current custody practices by examining the rights and obligations that custodians agree with their clients. To the best of our knowledge, the article represents the most comprehensive dataset of contract terms used by prominent crypto custodians.2

Contractual clauses were at the heart of large-scale crypto insolvencies and asset misappropriations of crypto firms culminating in the bankruptcy of FTX, now dubbed the ‘Crypto Winter’.3 The significance of crypto custodians’ T&Cs was highlighted in the Celsius Case4 in which Judge Glenn ruled that Celsius’s T&Cs ‘formed a valid, enforceable contract between [Celsius] and Account Holders.’ Conversely, he also found that those T&Cs unambiguously transferred ownership of the earned assets to the debtors, as account holders ‘grant Celsius … all rights and title to such Digital Assets, including ownership rights’.

The Crypto Winter events were not unprecedented; earlier examples date back to Mt Gox in 2014.5 The insolvencies in 2022–23 undermined investor trust in crypto-asset service providers (CASPs) and increased demand for custody solutions that safeguard clients’ assets effectively. These events also elevated crypto on the regulatory agenda.

While the crypto market recovered from the Crypto Winter and Bitcoin has experienced record highs, we note a strong trend towards regulation. Specifically, the EU legislator has introduced rules on crypto custodians within the framework of the EU Markets in Crypto-Assets (MiCA) Regulation,6 while the safekeeping of cryptos that fall under the financial instrument’s definition are covered by traditional rules on custody and safekeeping of financial instruments laid down in the MiFID.7 EU regulators have refrained from admitting crypto-assets8 as eligible assets for the EU retail fund brand, the UCITS.9

In the US, we observe several approaches to custodial holdings of cryptocurrencies: namely express trust, constructive trust, financial assets governed by Article 8 of the Uniform Commercial Code, bailment, and property sold to the intermediary.10 A SEC bulletin11 has sparked significant debate regarding whether it would be appropriate for financial institutions to report custodied crypto-assets as a liability on their balance sheets. While this Staff Accounting Bulletin 121 (SAB 121) is still in place, the US Securities and Exchange Commission appears lately to have adopted a more lenient stance, creating opportunities for new crypto custody and wallet providers.12

We are also witnessing the emergence of new crypto exchange-traded funds (ETFs);13 stablecoins such as those introduced by global payment service provider PayPal;14 Twitter/X’s announcement that it intends to enter into crypto and payments services;15 exchange-traded products based on Bitcoin and Ether; and TradFi-based crypto asset and custody solutions.16 This ongoing intermediation and institutionalization of crypto requires a sophisticated crypto custody framework. We expect international financial institutions to influence custodial practices,17 which already seems to be the case with respect to Blackrock and the amendment to the prime brokerage agreement with Coinbase, the custodian of its crypto ETFs.18

Drawing on a hand-selected dataset comprising 110 crypto custodians19 (from 48 jurisdictions)20 that made 88 T&Cs available on their websites in the period of May 2023–2024,21 we provide an overview of crypto custody practices identifying trends and patterns, as well as contradictions and divergences across the crypto industry. In doing so, we seek to ascertain whether the T&Cs of crypto custodians adequately protect their clients’ assets, particularly given that prospective clients accept standardized terms. With limited opportunities for direct communication and with portions of these terms being technically embedded, clients often have no means to negotiate. We also seek to ascertain what kind of legal framework is required to ensure a minimum level of protection for crypto clients in the context of custody services. As a regulatory benchmark, we use the custody requirements contained within MiCA.22

Following this introduction, section II provides an overview of how crypto-assets are held in custody via cryptographic key control. Section III introduces our research sample, while section IV discusses safekeeping, key storage transparency, outsourcing, omnibus wallets, asset segregation and insolvency protection, client entitlements, reuse, liability cap, and the choice of law, courts and arbitration. Section V argues and section VI concludes.

II. CUSTODY OF CRYPTO-ASSETS: WALLETS, KEYS, AND ADDRESSES

Unlike traditional assets, crypto-assets do not exist in tangible form. Instead, they constitute digital representations of a value or rights, and only exist due to the storage of data points by way of a given distributed ledger technology (DLT).23 Crypto-asset creation and transactions take place exclusively digitally, and are recorded and stored on the relevant DLT. The digital nature of crypto-assets contrasts with traditional notion of custody, which involves physically controlling and retrieving an asset. However, the apparent need to manage and safeguard rights and economic positions in crypto-assets has led to the development of alternative approaches being taken to crypto custody.

In a peer-to-peer network where all transactions are publicly announced,24 cryptography and the applicable encryption techniques provide the technical means to access and control the crypto-assets. The most prominent technique is public key encryption, which enables individuals, or service providers on behalf of their individual clients, to access their crypto-assets on the blockchain and exercise exclusive control through the use of keys, addresses, and wallets.25

1. Asymmetric/public key encryption

Asymmetric, or ‘public key’ encryption is a cryptographic technique invented in the 1970s that enables parties completely unknown to one another to share information over an unsecured connection. It consists of a mathematical function that generates a pair of asymmetric keys: one public, and the other private.26 Both private and public keys are needed to perform any task. Data are encrypted by the sender using the recipient’s public key and decrypted with the recipient’s private key.27 In cases of digital signatures,28 data are encrypted using the sender’s private key and decrypted with the sender’s public key.29

2. Crypto-assets: public-private key pair

In the crypto-asset ecosystem, public key encryption is vital to controlling access to crypto-assets and enabling asset transactions. Each user has at least one public key and one private key.30 The public key can be compared to a bank account number, while the private key is similar to a PIN code or signature on a bank cheque, ensuring that the holder enjoys exclusive control over the account balance.31

a. Public key and address

The public key is openly disclosed and cryptographically generated from its corresponding private key. It is a string of alphanumeric characters and serves as a unique identifier on the blockchain, acting as a pseudonym linked to a particular user to whom crypto-assets can be assigned, similar to a bank account number.32

The user wallet’s address on the blockchain is usually generated from its public key and this address functions similarly to the beneficiary’s bank account details in traditional finance. Anyone can send crypto-assets to a particular blockchain address or receive them, but only the user who controls the paired private key can actually make use of the crypto-assets assigned to it.

b. Custody as a means of safekeeping or control of private keys

The private key is the secret partner of the public key, and it is a randomly generated number. While the public key can be cryptographically generated from the private key, the opposite is mathematically impossible. The private key’s main function is to enable its holder to digitally sign a transaction spending x crypto-assets, thereby proving ownership of the assets involved in the transaction.33 While anyone can send crypto-assets to a wallet address on the blockchain, thereby assigning crypto-assets to a public key, only the one controlling the corresponding private cryptographic key has exclusive access to and control over the respective crypto-assets by technically proving ownership. The loss of the private key results in the irrevocable loss of the crypto-assets associated with the paired public key.34

This reality led the EU legislator to stipulate that the ‘custody and administration of crypto-assets’ includes the safekeeping or control of the crypto-assets, or of the means of access to such crypto-assets in the form of private cryptographic keys, on behalf of clients.35

3. Crypto wallets and wallet providers

Users rarely have direct knowledge of their own cryptographic keys. These keys are typically stored inside a crypto wallet, which is often operated by a crypto wallet provider.

a. Crypto wallets

A crypto wallet stores the private cryptographic keys that enable the client to hold, store, and transact in crypto-assets.36 There are several types of crypto wallets.37 A prominent distinction here is between hardware and software wallets, which are determined according to the nature of the wallet (ie a physical device like a USB stick (hardware) or a software application (software)). There are also hot, cold, or warm wallets, which are distinguishable by their use of online, offline, or combined methods of key storage for private keys.

b. Custodial and non-custodial wallet providers

The most significant distinction lies between custodial and non-custodial wallets,38 as this determines who controls the private key.39 Custodial wallet providers typically generate new public-private key pairs for the crypto-assets they hold in custody, rather than accepting a private key generated by another party, such as the client. This arrangement allows the intermediary to ensure it is the sole entity responsible for safeguarding and controlling the private cryptographic key (and, by extension, the associated crypto-assets). In contrast, non-custodial wallet providers claim to enable clients to hold their private keys directly. If accurate, this would effectively result in a form of self-custody by the holder.

III. ANALYTICAL APPROACH AND DATA SAMPLE

This study focuses on custodial wallet providers, and asks whether custody practice, as depicted in the T&Cs of prominent crypto custodians, is designed to protect their clients’ assets adequately.

For that purpose, we collected and analysed a sample of T&Cs used by a large number of custodial wallet providers, including all prominent crypto custodians. We then scrutinized the sample to ascertain what rights and positions they grant to clients.

1. Approach

Our research aims to understand how crypto custody providers exercise contractual freedom. Focal points include which services these providers offer, what happens in the case of insolvency, and the extent to which providers assume liability.40

For that purpose, we collected a sample of 88 T&Cs made available online on the websites of prominent custody providers in the period between May 2023–2024. Due to the criteria for their selection, we believe this sample to be representative of the overall market. We then analysed and summarized the insights gleaned therefrom, applying pre-set criteria to produce descriptive statistics.

2. Sample composition

Reliable estimates on the number of crypto custodians and their importance, measured by assets under custody or client numbers, are lacking. Our sample includes 88 T&Cs of 110 custody providers from 48 jurisdictions.

a. Selection criteria

The composition of the sample follows three criteria.

First, wallet providers are not legally required to disclose the crypto-assets held in their wallets, despite calls for greater transparency.41 Only a limited number of wallet providers, in particular certain leading centralized exchanges (CEXs), disclose on a voluntary basis the assets held in custody. Our sample of 110 providers includes the largest 20 providers that disclose their holdings.42

Second, CEXs are important custody providers, given that they offer custodial wallets to their clients to facilitate trading activities. Of the sample of 110 providers, 70 are leading CEXs,43 among which 20 CEX disclose their holdings, thus overlapping with the first category.

The third criterion is qualitative in nature and seeks to include custody providers that are not leading CEXs but are more well-known for the provision of custody services. For this, we searched for websites listing ‘top crypto custody providers’ and similar rankings. Of the 110 providers selected, 40 were listed in these rankings.44

b. Sample vs size of the market

Based on their proof of reserves (PoR) and similar rare disclosure of assets under custody or financial reserves,45 those that disclose these data among the selected custodians hold a share of at least 9.03% of the entire crypto-asset market by USD value.46  Figure 1 lists the top 10 CEX custodians by assets under custody (AuC) disclosed in our sample, revealing significant market concentration (Figure 2), with Binance being by far the most dominant provider.

Top 10 CEX Custodians by AuC disclosed. Data on the financial reserves of the custodians as of CoinMarketCap 21 March 2024.
Figure 1

Top 10 CEX Custodians by AuC disclosed. Data on the financial reserves of the custodians as of CoinMarketCap 21 March 2024.

Number of CEXs by AuC disclosed. Data on the financial reserves of the custodians as of CoinMarketCap 21 March 2024.
Figure 2

Number of CEXs by AuC disclosed. Data on the financial reserves of the custodians as of CoinMarketCap 21 March 2024.

According to the disclosed data on AuC, Binance holds a significantly larger amount in custody than all other CEX custodians combined. However, comparable data on AuC held in wallets provided by prominent crypto custodians—selected based on qualitative criteria—are entirely absent.

Noting an overall shortage of reliable disclosures for estimating market size, we believe our selection methodology ensures that most globally significant custody providers have been considered, offering a representative overview of market practice.

c. Geographical breakdown

For our sample comprising 88 T&Cs we relied on custodians that disclose the T&Cs on their websites. In Figure 3, the jurisdictions of their registered seat and/or crypto-related licence or registration are outlined as per the custodians’ website disclosures. As many custodians are part of groups with multiple entities operating or licensed across several jurisdictions simultaneously, our geographical breakdown includes all jurisdictions identified through their websites. In turn, the 88 T&Cs correspond to sample custodians that have, in total, 197 registered entities and/or crypto licences. It is often unclear to which entity a given licence belongs, or whether a given entity is licensed. Yet both the registered seat and the licence indicate the regional scope of activities. For that reason, we find it illuminating that 50 corporate seats and/or licences come from the EU; 23 from the rest of Europe; 43 from North and Central America; 47 from the Asia-Pacific region; 7 from the Middle East; 15 from Africa; eight from the Caribbean; and four from South America.

Breakdown of sample by region.
Figure 3

Breakdown of sample by region.

Figure 4 displays the top jurisdictions by the same measure (registered seats and/or licences) as follows: the US (31); Singapore (12); Lithuania (12); Australia (11); Switzerland (10); UK (9); Seychelles (9); Canada (8); Poland and Hong Kong (6 each); and Italy, France (5 each), Germany, BVI, Cayman Islands, Indonesia, South Korea (4 each).

Breakdown of sample by jurisdiction.
Figure 4

Breakdown of sample by jurisdiction.

IV. RESULTS

The results taken from our dataset allow us to make the following comments on crypto custody practices.

1. Safekeeping

The notion of safekeeping plays a pivotal role in financial legislation related to the custody of financial instruments. While there is no uniform definition of what constitutes the safekeeping of assets, an exhaustive list of safekeeping duties that a custodian is to assume is typically provided in the relevant legislation.47

The common denominator of safekeeping duties in financial legislation is that, for financial assets eligible for custody, the core principle—and conditio sine qua non—of safekeeping lies in the custodian holding those assets under its name on behalf of the client. Safekeeping traditionally encompasses the following two distinct cases: (i) for tangible assets that can be physically delivered to the custodian such as a painting or a piece of art, the custodian must be in a position to physically control and retrieve the assets; and (ii) for financial instruments that can be registered in a financial instruments account, the custodian must properly register those in such an account opened in the custodian’s books under the name of the client.

Crypto-asset practices challenge traditional custody by focusing on controlling access via private cryptographic keys. For instance, the MiCA rules on crypto custody—somewhat ambiguously—initially link safekeeping to the custody of crypto-assets, which subsequently encompasses the safeguarding of clients’ rights and the prevention of their reuse.48

We attempted to ascertain whether the service providers in our sample T&Cs explicitly assume crypto-asset safekeeping duties. This involves assessing whether they state in their terms that they offer a custody solution for their clients’ crypto-assets, provide (custodial) wallets, safeguard, safekeep assets, somehow hold, store, or control the client’s assets and/or the cryptographic keys, or use similar expressions.

Our sample reveals in Figure 5 that only 62 (71%) of our sample T&Cs mention that they offer a type of custodial, safekeeping service related to their clients’ crypto-assets. Meanwhile, 23 (26%) lack any relevant reference,49 while three (3%) deny50 the provision of a wallet service and/or custody to their clients.

Safekeeping.
Figure 5

Safekeeping.

We observe that well-established crypto custodians often avoid committing to the same custodial safekeeping obligations in their legally binding documents that custodians of financial instruments typically accept, instead opting for less explicit commitments. This is because there is either ambiguity about what crypto custody and safekeeping entails, or providers seek to optimize their legal position by avoiding commitments.

2. Key storage

The Crypto Winter demonstrated that custodians, especially exchange-linked wallet providers, prefer to use the so-called ‘hot storage’ of private keys of multiple clients to gain liquidity for short-term transactions.51 It is implied with the use of hot wallets that the cryptographic keys associated with the clients’ assets are generated and stored online, for instance on the custodian’s servers. These key storage tools are permanently connected online, ensuring time and cost efficiency. Conversely, hot wallets have been referred to as ‘an attractive “honey pot” for hackers’52 and are considered technically vulnerable.53 Some hot wallet providers employ security measures in the form of two-factor authentication and/or multi-signature support54 etc, but the cold wallets are still usually considered to provide more safety. Cold wallets, on the other hand, generate and store the private keys offline, such as on a USB stick, and are considered a safer variant of crypto custody.55 From a client’s perspective, knowing whether their assets are being held in hot or cold storage is crucial. Similarly, article 75(1) MiCA mandates crypto custodians to disclose their custody policies and the risks associated therewith in their custody agreement.

Against this background, we find in Figure 6 that an astonishing 69 (79%) of our sample T&Cs omit any reference to hot or cold storage, and one (1%) admitted to hot storage. Only two (2%) admit to using a cold storage mechanism while 16 (18%) employ a combined hot and cold storage mechanism.

Key storage transparency.
Figure 6

Key storage transparency.

3. Asset segregation, omnibus wallets, and outsourcing to third parties

Custodial arrangements for financial assets fall into two categories: omnibus or individually segregated accounts. Individual accounts show each client’s individual rights, while omnibus accounts present all clients’ assets collectively in one account. In principle, all accounts should be segregated and outline only an individual client’s assets. However, from the perspective of custodians, omnibus accounts are more cost efficient and easier to operate as they safeguard only one common pool of assets, while managing the rights of all clients in a register separate from the safeguarded assets. While custodians benefit from this arrangement, the same is not necessarily the case for clients. In other words, when all assets are pooled in an omnibus account, clients’ commingled entitlements are exposed to the risk of any other client’s insolvency. These entitlements also rely heavily on the custodian’s bookkeeping, which must be accurate and considered sufficiently robust by an insolvency administrator to serve as reliable evidence of clients’ entitlements in the event of the custodian’s insolvency.

Traditional finance (TradFi) regulation balances cost and client protection considerations. A custodian at the first level of the custody chain, ie a custodian entering into a contractual relationship with the client, is legally obliged to provide a fully segregated account for each of its clients individually and to properly record the assets of each client in its books and registers.56 This two-pronged obligation ensures the ring-fencing of each client’s assets against insolvency risks.

TradFi regulation further provides that if the custodian outsources safekeeping to a third party, the latter is entitled to operate omnibus accounts. This commingling at the second level is a technical necessity to avoid the undesirable scenario in which, within a hierarchical custody chain, all client data must be forwarded to, and reconciled with, the delegate. Even then, an omnibus account operated by the third party must exclude the proprietary assets of the delegating custodian, the proprietary assets of the third party, and the assets of the third party’s other clients. Notably, an omnibus account in TradFi is always a collective account of assets held on behalf of clients only; the assets of custodians and of any third-party safekeeping, as the case may be, are excluded. Secondly, the custodian and any third-party safekeeping must maintain detailed records and books showing segregation, ensuring they can promptly and accurately identify the assets belonging to each individual client at any time.57

For on-chain storage of crypto-assets, assets belonging to multiple users can either be segregated by assigning each user’s assets to an individually segregated wallet address on the blockchain or pooled into a collective or omnibus wallet address.58 The EU’s MiCA59 now imposes rules for crypto omnibus wallets similar to those governing financial instruments’ omnibus accounts: commingling multiple clients’ crypto-assets in omnibus wallet addresses is permissible. However, the crypto-assets of any custodian must be held separate from its clients’ assets ‘on the distributed ledger’60 as well as ‘operationally’ and ‘legally’.61 In other words, an omnibus wallet is recognized when it commingles the crypto-assets of multiple clients, while it is considered a (non-permitted) ‘collective’ wallet address when it mixes (multiple) clients’ assets with the custodian’s or any (sub)custodian’s own assets.

Omnibus wallets can leave a client exposed to risks stemming from the custodian’s other clients, whose assets are pooled inside the same wallet.62 In the event of a wallet hack or the insolvency of client B, the loss allocation and the attribution of the commingled assets between clients A, B, and C will depend on their agreed entitlements. Client A’s entitlement over the crypto-assets—whatever its legal form, depending on the forum and the applicable law63—will depend on whether the entitlement is (a) properly accounted for in the custodian’s accounts, and (b) recognizable and separable from other crypto-assets. In principle, the custodian’s properly maintained internal records may serve as evidence of the client’s entitlement. However, in many crypto insolvencies, the prerequisite for this legal presumption—orderly bookkeeping—has not been met.64

While on-chain storage can either have segregated or omnibus/collective wallet addresses, it is a legal necessity to segregate assets off-chain as well to identify which assets belong to each user at any time. However, there is ambiguity as to what off-chain segregation constitutes here. As per the International Swaps and Derivatives Association (ISDA),65 and what is derived from our sample, off-chain segregation ensures that the custodian maintains in its books and records a segregated account for each client, therefore ensuring the proper record-keeping of each client’s assets at any time via its accounting ledger.

While the segregated book-keeping obligation is essential in all cases, its efficiency in protecting clients hinges on two qualifications. Firstly, special attention must be paid to cases where omnibus or collective accounts66 are maintained in the books and records of the custodian and/or of the third-party safekeeping the assets. This is similar to what is found in TradFi, whereby we assume that the provision of separate accounts per client off-chain in their books and records is a logical necessity for custodians so to identify the account balance for each client. However, from the perspective of client protection, we consider that the reference in the T&Cs to a client account and the record keeping of the account balance is of limited significance. This is because the client account may also form part of an omnibus client account at the custodian’s or any third-party safekeeping’s level, thereby posing the same risks as analysed above.

Secondly, where custodians claim to practise ‘off-chain segregation,’ another interpretation may be that the custodian stores the cryptographic keys in a segregated manner.67 The latter is deductive for anyone familiar with TradFi custody, yet less so when only one set of private keys exists for the entire omnibus wallet.

Our research on the off-chain treatment of assets focuses on cases where custodians admit in their T&Cs to any commingling taking place off-chain by means of omnibus and/or collective accounts for multiple clients. Common key storage will also be discussed further.

Client protection rests on (a) who holds the asset on behalf of clients (custodian of choice vs any third-party safekeeping), (b) what the custodian segregates or commingles on-chain in the wallet addresses, and (c) what the custodian commingles off-chain in the accounts. In this regard, the T&Cs provide useful insights.

a. Outsourcing to third parties

As seen in Figure 7, in our sample, 64 T&Cs (73%) make no statement on outsourcing, while only 24 (27%) provide that appointing a third party for the safekeeping of the clients’ assets is permissible;68 and among those only four provide the name of the third party, while 20 provide no specification. Out of the 24 service providers that entitle themselves to appoint a third party for safekeeping, eight inform their clients that there may be no sufficient protection in case of insolvency of this third party, as discussed below.

T&Cs on outsourcing to third-party safekeeping.
Figure 7

T&Cs on outsourcing to third-party safekeeping.

On-chain commingling/segregation.
Figure 8

On-chain commingling/segregation.

In contrast to TradFi, the less hierarchical structure of DLTs may account for the relatively lower numbers observed here. Rather than being compelled to participate indirectly through outsourcing within a hierarchical custodial chain, DLTs enable all crypto custodians to function as nodes or asset holders.

b. On-chain commingling/segregation (wallet addresses)

Regarding on-chain commingling, as seen in Figure 8, eight of our sample T&Cs refer to a type of shared wallet address without specifying whether theses addresses commingle only clients’ assets (omnibus wallet address) or whether they also include the custodian’s own assets (collective wallet address).69 In that sense, seven T&Cs refer to collective wallet addresses.70 Meanwhile, three T&Cs stipulate that they provide omnibus wallet addresses for their clients’ assets only. At the same time, five T&Cs state that they provide both commingling and segregation: on the one hand, some form of omnibus or collective wallet address and, on the other, a segregated wallet address per client. One service provider mentions in its T&Cs the provision of segregated addresses per client. Finally, 64 omit any reference to on-chain commingling or segregation, as explained above.

c. Off-chain commingling (accounts and/key storage)

We now ask in Figure 9 whether custodians specify in their T&Cs that they commingle clients’ assets and entitlements off-chain through the use of omnibus accounts—which combine multiple clients’ assets—or collective accounts—which mix the custodian’s and/or any subcustodian’s own assets with the clients’ assets. Five T&Cs describe such collective accounts and seven T&Cs omnibus accounts accordingly. One set of T&Cs mentions that the keys to the assets of clients are stored collectively together with the keys to the custodian’s own assets. Finally, 75 T&Cs omit any reference to any off-chain commingling.71

Off-chain commingling.
Figure 9

Off-chain commingling.

4. Insolvency clauses and clients’ entitlements

The nature of clients’ entitlement over crypto-assets held in custody is a key factor determining the former’s position in insolvency.72 On the one hand, the client may have a proprietary right over the assets, and we may see this position blurred by an arrangement similar to traditional intermediated custody where the client retains the beneficial right only, and the custodian holds the legal right over the assets. On the other hand, the client may only have an (unsecured) contractual claim to receive equivalent assets based on a creditor-debtor relationship.

The holder of proprietary rights is protected in the event of insolvency because their crypto-assets will typically73 not be used to satisfy the custodian’s creditors. By contrast, the client-claimant is an unsecured creditor in the event of the custodian’s insolvency, similar to the custodian’s other creditors. If the claims against the insolvent estate exceed the assets, the assets of these clients will lose some or all of their value.74 The legal position taken may also depend on the respective third-party safekeeping arrangement, as it could vary between the two different approaches.75 At the first level, while the holding of the asset may come with a proprietary right, it may be that the law applicable to the third-party safekeeping arrangement results in an unsecured claim by the one relying on the third party (that is, contractually, the first-level custodian). As a result, proprietary rights at the first level may be lost, at least economically, in the event of the third party’s insolvency.

The above represent only some of the pitfalls relating to the legal entitlement of the asset-holders. Given that crypto property and insolvency laws are still largely undeveloped in most jurisdictions,76 the content of the T&Cs is crucial.

a. Clients’ entitlements to the crypto-assets

As seen in Figure 10, wording referring to the proprietary rights, title and ownership of clients relating to the custodied assets is mentioned explicitly in 17 T&Cs, while two T&Cs deny clients’ proprietary rights, with 69 T&Cs lacking any such stipulation. Beneficial ownership and trust arrangements for the custodied assets are mentioned in 10 T&Cs, while 78 T&Cs lack any stipulation thereon. Finally, eight T&Cs indicate explicitly that the client may have an (unsecured) claim in relation to the custodied assets against the custodian and/or any third-party safekeeping, while 80 T&Cs omit any stipulation on this matter.

Clients’ entitlements.
Figure 10

Clients’ entitlements.

Given its importance in determining the protection of assets, the large number of T&Cs lacking detailed stipulations is surprising. While most service providers use wording on their websites suggesting that clients ‘own’ their assets, their legal agreements tell quite a different story. Only 17 T&Cs mention that the title and ownership of the crypto-assets remain with the client, and just 10 mention beneficial holdings. In turn, depositing assets with the crypto custodians in our sample brings uncertainty as to clients’ entitlements, even impairing them outright. Given the cross-border nature of many client relationships and the associated conflicts of law therein, we find this situation worrying.

For instance, we have encountered clauses of CEXs stating that the client assigns all of their rights to all on-exchange assets to the CEX and agrees that their rights in relation to any on-exchange asset are limited to a contractual obligation against the custodian.77 Furthermore, even if the client is mentioned as having a proprietary right, in five T&Cs it is still stated that the client bears the risk of insolvency of the custodian or the third-party safekeeping the assets.

Even when T&Cs state that title and property remain with the client, this protection depends on other clauses, such as asset segregation, prohibition of reuse, and outsourcing arrangements. If the other clauses do not move conceptually in the same direction, a mere reference stating that title and property remain with the client, or similar wording to that effect, may become a vague expression that is contradicted by the custodian’s overall practice.

b. Insolvency protection clause

As seen in Figure 11, in the event of the custodian’s insolvency, six T&Cs stipulate that their clients’ crypto-assets are protected, and six T&Cs state that there may be no such protection—insolvency may lead to total loss of the clients’ assets and/or the clients may only have an unsecured claim. One set of T&Cs provides two different wallet types: one ensuring protection in case of insolvency, and the other not. Finally, 75 T&Cs are silent on the matter.

Insolvency protection.
Figure 11

Insolvency protection.

In the event of the insolvency of the third-party safekeeping, no T&Cs in our sample stipulate that the clients’ assets are protected, while eight mention that such protection may be lacking—the insolvency may lead to total loss of the clients’ assets and/or the clients may only have an unsecured claim. Notably, 80 T&Cs remain silent on the issue.

The very essence of custody-as-a-service with its emphasis on safekeeping appears fundamentally at odds with the situation presented in the observed T&Cs.

5. Right of reuse

With TradFi, clients’ assets held in custody must not be reused by the custodian for its own account or by any third party responsible for safekeeping. Reuse includes any transaction involving the custodied assets, including but not limited to transferring, pledging, selling, lending, or investing.78 The rationale applied here is to avoid assets escaping from the custodian’s exclusive sphere of control, thereby mitigating the risk of the custodian being unable to safekeep and deliver them when required. The financial regulation79 typically provides for an exception to the prohibition of the reuse of the financial assets on the premise that the custodian is transparent, there is prior explicit consent of the client, and in certain cases there is even the provision of collateral of an equivalent amount by the custodian. Asset reuse has been cited as a reason for asset loss in many recent crypto insolvencies.80

We find it surprising that, in Figure 12, 74 (84%) T&Cs remain silent on the issue, while only 12 (14%) rule out the reuse of the clients’ assets for their own benefit and purposes, with two (2%) even allowing for such reuse.

Right to reuse clients’ assets.
Figure 12

Right to reuse clients’ assets.

Liability cap up to the market value of any lost assets.
Figure 13

Liability cap up to the market value of any lost assets.

Among those that permit reuse, this encompasses any utilization of clients’ assets for the service provider’s own investments, liquidity provision, operational expenses, or other corporate or business purposes. These clauses do not come with obligations to put in place any other legal safeguards such as explicit consent and/or the provision of collateral. Surprisingly, out of 17 T&Cs that refer to the proprietary rights of clients on the custodied assets, only seven mention that the custodian will not reuse those assets, with one even allowing for such reuse.

The client’s position would clearly benefit from more transparency on the right to reuse assets.

6. Liability cap

Central to the custodian’s function is the obligation to safekeep clients’ assets, coupled with the duty, in the event of asset loss, to return assets of an identical type or an equivalent value represented by the crypto-asset. In the context of the crypto-asset custody, liability of the crypto custodian arises when the crypto-asset or the private keys are lost, whereby the custodian is obliged to indemnify the client up to the market value of the crypto-asset that was lost, at the time the loss occurred.81

As seen in Figure 13, at odds with the core function of custody, 44 (50%) of our sample cap their liability up to the total amount of fees that the client has paid to them.82 Ten (11%) T&Cs cap their liability at the market value of the lost crypto-assets, while 34 (39%) remain silent on the matter.

In the aftermath of the Crypto Winter and the corresponding massive losses in asset value, a liability cap for custodians significantly below the value of the actual damage suffered by clients renders the custodian’s safekeeping obligation meaningless.

7. Choice of law, courts, and arbitration clause

In the absence of robust private international rules applicable to the idiosyncrasies of crypto-assets and the difficulties in locating the CASPs, all service providers must provide information to their clients, prospective or otherwise, regarding the applicable law on the custodial relationship and the jurisdiction or arbitration process where the clients can exercise their legal rights.

Overall, as seen in Figure 14, 74 (84.09%) of the T&Cs provide a choice of law clause, but this is absent in 14 (15.91%) of them. A courts clause exists in 48 (54.55%) of our sample, and is omitted in 40 (45.45%). Meanwhile, 48 (54.55%) of the T&Cs provide an arbitration clause, while 40 (45.45%) omit any such reference.

Clause on choice of law, courts, and arbitration.
Figure 14

Clause on choice of law, courts, and arbitration.

As for choice of law clauses in Figure 15, the most common jurisdictions identified were the US (18 T&Cs), the UK (13), Singapore (8), Switzerland (6), and Hong Kong (4).

Choice of law.
Figure 15

Choice of law.

In addition, in Figure 16, we identified that the prominent fora identified in the T&Cs are US federal states (12), Switzerland (6), the UK (5), Singapore (4), and South Korea (3).

Choice of courts.
Figure 16

Choice of courts.

V. THE NEED FOR REGULATION OF CRYPTO CUSTODY

1. The state of play

According to our findings, most custodians are unwilling to legally and explicitly assume responsibility when safekeeping their clients’ assets; some even hesitate to publish their T&Cs on their website despite the benefits of attracting new clients in doing so. Many crypto custodians use misleading terminology, thus blurring the true extent of their accountability and liability with respect to their services. The lack of transparency regarding custody policies, particularly concerning the storage mechanism, is widespread in collective custody, carrying serious risks for investors.83 When discussing asset segregation and collective custody, our observations highlight legal uncertainties surrounding these concepts which result in inconsistent practices or practices that are inaccurately represented. In certain cases, the clients’ legal entitlements vis-à-vis their crypto-assets deteriorate when clients appoint a crypto custodian, and even more so when safekeeping is appointed to a third party. The result is uncertainty regarding the clients’ protection, especially in the event of the insolvency of the custodian or a third-party responsible for safekeeping.

The language found in the reviewed T&Cs regarding the right to reuse clients’ assets for custodians’ own business or investment purposes lacks the safeguards typically required under TradFi legislation, thereby increasing the risks borne by clients. The custodians in our sample tend to cap their liability not up to the value of clients’ assets lost, but only up to the amount of fees paid to them. Despite being crucial, only a few custodians provide guidance to their clients on the applicable law, jurisdiction, and arbitration process.

2. A market failure?

Our findings reveal a stark dichotomy in the crypto custody industry: while certain custodians prioritize client rights and interests, many more offer terms and conditions that conflict with meaningful client protection. This insight confirms earlier observations by Awrey and van Zwieten that alternative value transfer mechanisms, including centralized crypto exchanges, very rarely combine legal mechanisms that together ‘would come very close to replicating the unique protections typically afforded bank depositors’.84 Our sample data also showed that some of the larger providers’ T&Cs are particularly beneficial to the custodians.

The absence of private law-based protections in most cases arises from two key factors. The first contributing factor is that crypto custody serves as a necessary side-service for most retail and institutional crypto clients who seek exposure to crypto-assets but lack either the technical expertise for self-custody solutions or the investment volume to demand bespoke custody arrangements tailored to their needs. We find it notable that Fidelity, one of the world’s largest investment firms, developed their own trading and custody solution for their clients rather than relying on existing service providers when entering the crypto-assets investments market.85 Most clients lack the financial and technical expertise as well as the market power of firms like Fidelity, and are therefore compelled to rely on the crypto custody providers associated with the crypto trading platforms on which they trade.

The second contributing factor is the presence of liquidity shortages in many crypto-assets, paired with technical platform features resulting in economies of scale and scope as well as network effects. Crypto trading platforms, which constitute the largest part of our sample, often experience liquidity shortages, leading to significant bid-ask spreads and delayed transactions. Some platforms succeed in attracting substantial trading volumes in the most prominent crypto-assets.

As was laid out in the literature regarding two-sided platform markets,86 concerns other than legal particularities are prioritized when clients are deciding where to trade. In relation to marketplaces, liquidity is more important to clients than high-quality custody. Liquidity shortages have an immediate and certain price impact, whereas the downsides of less favourable custody terms are typically experienced only in specific cases and over the longer term. The combination of liquidity, data availability, and efficient trading technology collectively determines where trading volume is concentrated.

As long as the market for crypto custody is strongly linked to crypto exchanges, clients will be less concerned about quality custody, hence clients opting for a preferred crypto custodian will continue to play a less important role. Liquidity undermines any market-based adjustment of crypto custody T&Cs. Under these conditions, all but the most sophisticated, tech-savvy clients are unlikely to protect themselves through carefully chosen providers or sophisticated negotiations.

Crypto custody markets suffer from market failure, with large-scale externalities present. In turn, regulation is indispensable to ensure a proper level of consumer protection.87

3. Benefits of regulation

Where market failures exist, financial regulation can play an instrumental role in advancing—or may even be a crucial precondition for—a mature crypto industry. Where clients ‘generally enjoy relatively limited structural, contractual, or other private legal protections’ there is an ‘enormous pressure on public regulatory frameworks to ensure a sufficient level of consumer protection’.88 In the absence of such a framework, clients may lose trust in crypto custody, potentially leading them to exit the crypto asset market entirely or to avoid entering it with serious intent.

In such an environment, financial regulation can serve three purposes. First, it can promote a level playing field for crypto custodians where merit-based competition can develop and liquidity at crypto-exchanges plays a less important role.

Second, financial regulation can ensure adequate client protection, regardless of clients’ due diligence, and costly negotiations. This is all the more significant given that custody conditions are often embedded in technology, including smart contracts. Amending T&Cs often requires additional investment in adjusting technology. Take our example of hot and cold storage. A simple change of contractual language will be insufficient to improve client protection. Rather, the custody provider will be required to set up its system in an entirely new way and is unlikely to make further investments unless (a) all clients require the change at the same time (which is unlikely given the costs of coordination among large numbers of, often uninformed, crypto clients) and (b) competitors have to bear the same costs at the same time. Regulation can ensure that these conditions are provided.

Third, regulatory minimum standards of organization and conduct for custodians will reduce the likelihood of costly disputes, which may reduce information asymmetry, transaction costs, and potential losses for clients. Even when clients receive 100 cents on the dollar (as it appeared in the FTX insolvency89), regulation ensuring segregated accounts could assist in avoiding lengthy uncertainty on the value of clients’ accounts—in the case of FTX for 18 months. Regulation may not only prevent clients from experiencing the downsides of a crypto custodian’s insolvency, but also secure the potential upside of clients’ exposures—an outcome achievable only if client accounts are fully segregated, and the value of their assets is unequivocally attributed to them.

While it may be clear that regulation is beneficial, this does not, in itself, determine how regulation should be designed.90 Indeed, while the unbundling of trading and custody solutions may be welcomed as a means to unlock market forces and enhance customer choice in the crypto custody market, we must also acknowledge the new risks that arise from inconsistent technical integration of custody solutions into trading platforms. In such a technically complex environment, these risks could directly affect clients. Therefore, we urge regulators to proceed with caution.

Our sample suggests that regulators should focus on ascertaining a certain minimum content of contractual stipulations, including the name of the party, the applicable law, provisions on the choice of courts, arbitration clauses, and a potential remediation mechanism inside and outside of the insolvency. In addition, regulators should consider bespoke rules on asset segregation and safeguarding of clients’ assets, the clients’ rights (including assurance of propriety rights), storage methods, the reuse of assets, and outsourcing and liability caps, as well as fees charged. In this way, regulation could safeguard clients against the shortcomings arising from the underdeveloped private law framework governing crypto-assets.

VI. CONCLUSION

This article has provided an overview of the rights and obligations assigned by the T&Cs of crypto custodians to their clients.

To date, crypto has operated with minimal regulation and significant contractual freedom. This has resulted in a bifurcated industry. We observe some investor-oriented T&Cs, but also find a large sample of T&Cs that primarily serve to shield the crypto custodian from any meaningful accountability—contrary to what the words ‘safekeeping’ or ‘custody’ entail.

Financial regulation will be crucial to increase or maintain investors’ confidence in crypto by ensuring minimum protection of clients. Financial regulation could bring the benefit of merit-based competition across providers by ensuring a level playing field. Given that we could not demonstrate a relation between the estimated size of provider and client protection propensity, in the absence of mandatory custody regulation a ‘race to the bottom’ across the industry is not unlikely.

Footnotes

1

The global digital asset custody market was valued at USD 447.9 billion in 2022 and is expected to reach USD 1.601 trillion by 2028. Proficient Market Insights, ‘Digital Asset Custody Market’ (Globenewswire, 9 February 2023) <https://www.globenewswire.com/news-release/2023/02/09/2604673/0/en/Digital-Asset-Custody-Market-2023-will-Revenue-to-Cross-reaching-USD-1601115-31-million-by-2028-with-CAGR-of-23-65-during-the-forecast-period-Top-Companies-report-covers-Market-spe.html> accessed 13 December 2024.

2

See Dan Awrey and Kristin van Zwieten, ‘Mapping the Shadow Payment System’ (2018) 43 Journal of Corporation Law 775 (providing an empirical mapping of some of the questions discussed herein on a smaller scale and with a much broader concept of ‘shadow payment providers’; in turn, the results of the Awrey and van Zwieten study are less focused on crypto custody).

3

For a detailed analysis of the ‘Crypto Winter’ see Douglas Arner and others, ‘The Financialization of Crypto: Designing an International Regulatory Consensus’ (2024) 53 Computer Law & Security Review 105970.

4

Re Celsius Network LLC, 22-10964 (MG), [2023] Bankr LEXIS 2 (Bankr SDNY Jan 4, 2023); see Jonah Wacholder and Daniel Lowenthal, ‘Bankruptcy Court Issues Ruling on Ownership of Celsius Account Assets’ (Patterson Belknap, 31 January 2023) <https://www.pbwt.com/bankruptcy-update-blog/bankruptcy-court-issues-ruling-on-ownership-of-celsius-account-assets> accessed 13 December 2024; Arnold and Porter, ‘Read Before You Click “Accept”: Judge Glenn Rules That Earn Account Crypto Assets are Property of Celsius’ Bankruptcy Estates and Not Customer Property’ (Arnold & Porter, 5 January 2023) <https://www.arnoldporter.com/en/perspectives/advisories/2023/01/read-before-you-click-accept> accessed 13 December 2024.

5

See for an account of the Mt Gox insolvency, Awrey and van Zwieten (n 2) 796–99.

6

Regulation (EU) 2023/1114 of the European Parliament and of the Council of 31 May 2023 on markets in crypto-assets, and amending Regulations (EU) No 1093/2010 and (EU) No 1095/2010 and Directives 2013/36/EU and (EU) 2019/1937 (MiCA) [2023] OJ L150/40. On details, see Dirk Zetzsche, Julia Sinnig and Areti Nikolakopoulou, ‘Crypto Custody’ (2024) 19 Capital Markets Law Journal 207, 208–12.

7

Directive 2014/65/EU of the European Parliament and of the Council of 15 May 2014 on markets in financial instruments and amending Directive 2002/92/EC and Directive 2011/61/EU (recast) (MiFID) [2014] OJ L173/349.

8

For instance, a UCITS could gain exposure to crypto-assets by way of Exchange-Traded Products (ETPs) qualifying as security.

9

The European Commission has asked the European Securities and Markets Authority (ESMA) to review the UCITS Eligible Assets Directive 2007/16/EC, also to consider investments into digital assets. See European Commission, Formal request to ESMA for technical advice on the review of Commission Directive 2007/16/EC on UCITS eligible assets, Ref Ares(2023)3906255 – 06/06/2023; ESMA has launched a publication consultation in the second quarter of 2024. We expect the results to take into account the coming into force of MiCA on 1 January 2025.

10

On details see Adam J Levitin, ‘Not Your Keys, Not Your Coins: Unpriced Credit Risk in Cryptocurrency’ (2022) 101 Texas Law Review 877, 27ff <https://ssrn.com/abstract=4107019> accessed 13 December 2024.

11

See SEC, ‘Staff Accounting Bulletin No. 121’ (SEC, 31 March 2022) <https://www.sec.gov/regulation/staff-interpretations/accounting-bulletins/old/staff-accounting-bulletin-121> accessed 13 December 2024; David Stier, Eric Forni and Eric Hall, ‘The Saga of SAB 121: Losing a Battle, yet Gaining Ground’ (DLA Piper, 31 July 2024) <https://www.dlapiper.com/en/insights/publications/blockchain-and-digital-assets-news-and-trends/2024/the-saga-of-sab-121> accessed 13 December 2024.

12

Brian McGleenon, ‘How BNY Mellon’s SEC approval could reshape crypto custody landscape’ (Yahoo finance, 27 September 2024) <https://uk.finance.yahoo.com/news/bny-mellon-sec-approval-crypto-050028161.html> accessed 13 December 2024.

13

On 10 January 2014, after a string of successful lawsuits against the SEC reluctant stance, the SEC approved the conversion of Grayscale trust into an exchange-traded product. See Gary Gensler, Chairman of the SEC, Statement on the Approval of Spot Bitcoin Exchange-Traded Products (SEC, 10 January 2024) <https://www.sec.gov/newsroom/speeches-statements/gensler-statement-spot-bitcoin-011023> accessed 13 December 2024. Since 10 January 2024, for instance BlackRock, Fidelity and Invesco have gained approval by the US SEC for bitcoin exchange-traded funds; see Paul Katzeff, ‘SEC Approves New Bitcoin ETFs: What It Means for Investors’ (Forbes, 11 January 2024) <https://www.forbes.com/advisor/investing/cryptocurrency/sec-approves-new-bitcoin-etfs/> accessed 13 December 2024.

14

PayPal, ‘PayPal Stablecoin’ <https://www.paypal.com/us/digital-wallet/manage-money/crypto/pyusd> accessed 13 December 2024.

15

Sebastian Sinclair, ‘Twitter ‘X’ rebrand spurs hype around potential crypto integration—again’ (Blockworks, 26 July 2023) <https://blockworks.co/news/twitter-x-rebrand-crypto> accessed 13 December 2024.

16

For instance, BNY Mellon has obtained a ‘no-objection’ from the US Securities and Exchange Commission (SEC) regarding its request to safeguard digital assets such as bitcoin and ether without the need to classify them as balance-sheet liabilities. This will enable the bank to enter into the bitcoin and ether crypto-custody market, see Zacks Equity Research ‘BNY Mellon Receives SEC’s Approval to Offer ETF Crypto Custody Service’ (Yahoo finance, 30 September 2024) <https://finance.yahoo.com/news/bny-mellon-receives-secs-approval-150900804.html> accessed 13 December 2024. Fidelity Digital Assets currently offers bitcoin and ether custody and trading on its own platform, see Fidelity Digital Assets <https://www.fidelitydigitalassets.com/trading-custody> accessed 13 December 2024.

17

This observation is based on the custody arrangements offered by Coinbase Custody Trust Company LLC, which offers custody and wallet arrangements for most of the US SEC-approved bitcoin ETFs. A first look at the terms disclosed in the prospectus reveals significant variety of T&Cs. A detailed analysis of this variety would warrant a separate study focused specifically on US ETFs.

18

iShares Bitcoin Trust ETF, Form 8-K report (iShares, 16 September 2024) <https://www.ishares.com/us/literature/fund-announcement/ibit-8k-09-19-24.pdf> accessed 13 December 2024; SEC, ‘Amendment to the Coinbase prime broker agreement’, Exhibit 10.1 (SEC, 16 September 2024) <https://www.sec.gov/Archives/edgar/data/1980994/000143774924029619/ex_725171.htm> accessed 13 December 2024; see Bitget, ‘BlackRock Modifies Coinbase Custody Agreement to Require 12-Hour Withdrawals’ (Bitget, 20 September 2024) <https://www.bitget.com/news/detail/12560604219313> accessed 13 December 2024.

19

Out of 110 identified crypto custodians, 22 either do not provide a website or on May 2024 did not disclose T&Cs on their website, resulting in a sample of 88 T&Cs for this analysis.

20

We draw the geographical data exclusively from the websites of the custodians, ie we used the jurisdiction disclosed on their websites.

21

More than 50 per cent of the custodians of our sample provide several T&Cs, agreements and/or policies on their websites. This is particularly true, first, for providers offering multiple services—even services non-related to crypto assets—or, second, for providers active in multiple jurisdictions. In the first case, priority is given to any custody agreement for crypto assets addressed to the public (if any), supplemented by the General T&Cs; both together are counted as one agreement. However, given we often could not identify bespoke custody agreements (titled as such), in many cases, ie 83 (94.3 per cent of T&Cs), our sample provided the content of the general T&Cs and we emphasized on the respective clauses of custodial interest. Risk disclosures, privacy policies, and other special agreements such as for margin services or NFTs displayed on the website are out of the scope of our T&Cs analysis. In the second case, where there are multiple T&Cs because one custodian is active in multiple jurisdictions, we analyse one of the T&Cs, chosen on a discretionary basis.

22

Regulation (EU) 2023/1114 (n 6).

23

Satoshi Nakamoto, ‘Bitcoin: A Peer-to-Peer Electronic Cash System’ <https://bitcoin.org/bitcoin.pdf> accessed 13 December 2024.

24

Malte Moser, ‘Anonymity of bitcoin transactions’ (Münster Bitcoin Conference, 17–18 July 2013) 1–4 <https://allquantor.at/blockchainbib/pdf/moser2013anonymity.pdf> accessed 13 December 2024.

25

The EU’s MiCA recognizes this practice of accessing and controlling the crypto-assets on the blockchain via keys as a form of custody of crypto-assets. MiCA, article 3(1)(17) defines the service of ‘providing custody and administration of crypto-assets on behalf of clients’ as the ‘safekeeping or controlling, on behalf of clients, of crypto-assets or of the means of access to such crypto-assets, where applicable in the form of private cryptographic keys’.

26

Whitfield Diffie and ME Hellman, ‘New Directions in Cryptography’ (1976) 22 IEEE transactions on information theory 644, 647–49; Marco Conoscenti, Antonio Vetrò and Juan Carlos De Martin, ‘Blockchain for the Internet of Things: A Systematic Literature Review’ (2016) IEEE / ACS 13th International Conference of Computer Systems and Applications (AICCSA), Agadir, Morocco, 2 <https://ieeexplore.ieee.org/document/7945805> accessed 13 December 2024.

27

Elaine Barker and others, ‘Guide to IPsec VPNs’ (2020) NIST Special Publication 800-77 Revision 1, 143 <https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-77r1.pdf> accessed 13 December 2024; Jan Witte, ‘The Blockchain: A Gentle Introduction’ (2016) 2 <https://ssrn.com/abstract=2887567> accessed 13 December 2024; Joshua Kroll, IC Davey, and EW Felten, ‘The Economics of Bitcoin Mining, or Bitcoin in the Presence of Adversaries’ (2013) Twelfth Workshop on the Economics of Information Security (WEIS 2013) Washington, DC <https://api.semanticscholar.org/CorpusID:2794725> accessed 13 December 2024.

28

Carol Goforth and Yuliya Guseva, Regulation of Cryptoassets, American Casebook Series (2nd edn, West Academic Publishing 2022) 790; Ronald Rivest, Adi Shamir and Leonard Adleman, ‘A Method for Obtaining Digital Signatures and Public-Key Cryptosystems’ (1970) 21 ACM Communications 120 <https://dl.acm.org/doi/10.1145/359340.359342> accessed 13 December 2024.

29

Primavera De Filippi and Aaron Wright, Blockchain and the Law: The Rule of Code (Harvard University Press 2018) 14–16; Leighton Johnson, Security Controls Evaluation, Testing, and Assessment Handbook (2nd edn, England: Academic Press 2020) 524–26.

30

Each user can have as many key pairs as they wish and perform transactions under various pseudonyms, see Philipp Maume, Lena Maute and Mathias Fromberger, The Law of Crypto Assets (Bloomsbury 2022) 6–10.

31

Stephen Small, ‘Bitcoin: The Napster of Currency’ (2015) 37 Houston Journal of International Law 581, 588ff; Andreas Antonopoulos, Mastering Bitcoin: Programming the Open Blockchain (3rd edn, O’Reilly Media 2023) ch 4.

32

Ibid; Goforth and Guseva (n 28) 760–81, 790.

33

Inter alia Matthias Haentjens, Tycho de Graaf and Ilya Kokorin, ‘The Failed Hopes of Disintermediation: Crypto-Custodian Insolvency, Legal Risks and How to Avoid Them’ (2020) Singapore Journal of Legal Studies 526 <https://ssrn.com/abstract=3589381> accessed 13 December 2024; Wulf Kaal and Hayley Howe, ‘Custody of Digital Assets’ (2021) 5 University of St Thomas (Minnesota) Legal Studies Research Paper <https://ssrn.com/abstract=3936876> accessed 13 December 2024; Dominik Skauradszun and Jeremias Kümpel, ‘Crypto Custodians in Financial Distress’ (2023) 32 International Insolvency Review 538 <https://ssrn.com/abstract=4324529> accessed 13 December 2024; Hossein Nabilou, ‘The Law and Macroeconomics of Custody and Asset Segregation Rules: Defining the Perimeters of Crypto-Banking’ (2022) 9 Amsterdam Law School Research Paper; 3 Amsterdam Center for Law & Economics Working Paper <https://ssrn.com/abstract=4075020> accessed 13 December 2024.

34

Matthew Amlot, ‘Around $140 billion of Bitcoin locked away forever: Report’, (ALARABIYA, 13 January 2021) <https://english.alarabiya.net/business/technology/2021/01/13/Around-140-billion-of-bitcoin-locked-away-forever-Report> accessed 13 December 2024.

35

MiCA, art 3(1)(17).

36

De Filippi and Wright (n 29) 21; Small (n 31) 588; Maume, Maute and Fromberger (n 30) 10; Haentjens, de Graaf and Kokorin (n 33) 9.

37

Wenkai Li and others, ‘A Survey of DeFi Security: Challenges and Opportunities’ (2022) 34 Journal of King Saud University—Computer and Information Sciences 10378, 10382–10384 <https://www.sciencedirect.com/science/article/pii/S1319157822003792> accessed 13 December 2024; Hossein Rezaeighaleh and Cliff Zou, ‘New Secure Approach to Backup Cryptocurrency Wallets’ (2019) IEEE Global Communications Conference (GLOBECOM), Waikoloa, 1–6 <https://ieeexplore.ieee.org/document/9014007> accessed 13 December 2024.

38

For example, see Tom Josua Barbereau and Balázs Bodó, ‘Beyond Financial Regulation of Crypto-Asset Wallet Software: In Search of Secondary Liability’ (2023) 49 Computer Law & Security Review 105829.

39

For a taxonomy of the non-custodial and semi-custodial wallet solutions, see Zetzsche, Sinnig and Nikolakopoulou (n 6) 214ff.

40

Compared to Awrey and van Zwieten (n 2), our approach focuses on crypto custody rather than cryptocurrency trading. In turn, our scope of analysis is more granular.

41

After the ‘Crypto Winter turmoil’ and the subsequent bankruptcies that have taken place, in order to prove their solvency and re-establish investors’ trust some commentators demanded that the large centralized exchanges should promote transparency by disclosing the crypto-assets they hold on behalf of clients via the so-called ‘Proof of Reserves’ (PoR) and similar techniques. See Greg Hopper, ‘How Can Run Risk in Digital Asset Markets Be Reduced?’ (2023) 16 Journal of Risk Management in Financial Institutions 383; Konstantinos Chalkias, Panagiotis Chatzigiannis and Yan Ji, ‘Broken Proofs of Solvency in Blockchain Custodial Wallets and Exchanges’ (2022) 43 International Conference on Financial Cryptography and Data Security 106 <https://eprint.iacr.org/2022/043> accessed 13 December 2024; Gaby Dagher and others, ‘Provisions: Privacy-preserving Proofs of Solvency for Bitcoin Exchanges’ (2015) CCS ’15: Proceedings of the ACM Conference on Computer and Communications Security 720 <https://eprint.iacr.org/2015/1008.pdf> accessed 13 December 2024; Vitalik Buterin, ‘Why we need wide adoption of social recovery wallets’ (Vitalik, 11 January 2021) <https://vitalik.eth.limo/general/2021/01/11/recovery.html> accessed 13 December 2024.

42

The website used to identify assets held in crypto custody was primarily <https://coinmarketcap.com/>. Providers added applying size criteria include, for instance, Binance, KuCoin, OKX, MEXC, Bitfinex.

43

The website from which our sample of leading custodial CEXs was taken was primarily CoinMarketCap, ‘Top Cryptocurrency Spot Exchanges’ (Coinmarketcap) <https://coinmarketcap.com/rankings/exchanges/>. Providers added applying this criterion include, for instance, Coinbase, Kraken, Gate.io, Gemini, Bybit.

45

There is no uniform definition of proof of reserves, assets under custody, nor can we be certain that all proof of reserves are calculated in the same way by CEX or other crypto custody and wallet providers reported by CoinMarketCap. We assume for purposes of our dataset this.

46

Data on the entire crypto market industry as of CoinMarketCap 21 March 2024.

47

For the fund depositary’s safekeeping duties, see Directive 2011/61/EU of the European Parliament and of the Council of 8 June 2011 on Alternative Investment Fund Managers and amending Directives 2003/41/EC and 2009/65/EC and Regulations (EC) No 1060/2009 and (EU) No 1095/2010 (AIFMD) [2011] OJ L174/1, article 21(8); Directive 2009/65/EC of the European Parliament and of the Council of 13 July on the coordination of laws, regulations and administrative provisions relating to undertakings for collective investment in transferable securities (recast) (UCITSD) [2009] OJ L302/32, article 22(5); their respective delegated acts, in particular, Commission Delegated Regulation (EU) 2018/1618 of 12 July 2018 as regards the safe-keeping duties of depositaries [2018] OJ L271/1, rec 2; Commission Delegated Directive (EU) 2017/593 of 7 April 2016 with regard to safeguarding of financial instruments and funds belonging to clients, product governance obligations and the rules applicable to the provision or reception of fees, commissions or any monetary or non-monetary benefits [2016] OJ L87/500. For the MiFID custodian’s safekeeping duties, see MiFID, article 16, in conjunction with Commission Directive 2006/73/EC of 10 August 2006 implementing Directive 2004/39/EC of the European Parliament and of the Council as regards organisational requirements and operating conditions for investment firms and defined terms for the purposes of that Directive [2006] OJ L241/26, articles 16–20, and Commission Delegated Directive (EU) 2017/593, articles 2–8.

48

MiCA, article 70(1) in conjunction with article 75(3).

49

In the N/A category we also include three sample T&Cs that mention nothing regarding the custody service except a mere sentence that any provider’s custodial services are governed by a different agreement, which was not accessible to us. Furthermore, one T&Cs describing only a semi-custodial and/or non-custodial service where the service provider does not hold enough keys to perform a transaction without the client’s cooperation is also part of the N/A category. For our taxonomy on semi-custodial and self-custodial solutions, see Zetzsche, Sinnig and Nikolakopoulou (n 6) 214ff.

50

One out of those service providers denied wallet provision in its T&Cs but mentioned safekeeping in its custody statement.

51

Dirk Zetzsche and others, ‘Remaining Regulatory Challenges in Digital Finance and Crypto-Assets after MiCA’ (2023) Study for the European Parliament, 46–47 <https://ssrn.com/abstract=4487516> accessed 13 December 2024.

52

Parma Bains and others, ‘Regulating the Crypto Ecosystem: The Case of Unbacked Crypto Assets’ (September 2022) 7 Fintech Notes 22 <https://www.imf.org/en/Publications/fintech-notes/Issues/2022/09/26/Regulating-the-Crypto-Ecosystem-The-Case-of-Unbacked-Crypto-Assets-523715> accessed 13 December 2024.

53

For an analysis of the multiple hot wallet hacks of prominent custodians during the Crypto Winter, see Zetzsche, Sinnig and Nikolakopoulou (n 6) 208–10.

54

Forbes, ‘Crypto Hot Wallets vs Cold Wallets: The Key Differences’ (Forbes, 25 August 2024) <https://www.forbes.com/sites/digital-assets/article/crypto-hot-wallets-vs-cold-wallets/> accessed 13 December 2024.

55

See Antonopoulos (n 31) ch 13.

56

See AIFMD, article 11(d); UCITSD, article 22a(3); Commission Delegated Regulation (EU) 2018/1618, article 2(1)(d).

57

For the permissibility of fund depositary’s omnibus accounts see AIFMD, article 21(8) and (11); UCITSD, articles 22(5), 22a, and their respective delegated acts. For MiFID custodian’s omnibus accounts permissibility, see Commission Directive 2006/73/EC, article 16, in conjunction with Commission Delegated Directive (EU) 2017/593, article 2. See also ESMA Opinion, ‘Asset segregation and application of depositary delegation rules to CSDs’, ESMA34-45-277.

58

For the notion of wallet address see section II.2.a.

59

MiCA, article 75(7).

60

On a blockchain any crypto custodian needs at least two wallet addresses, one assigned to own crypto-assets and one to clients’ assets.

61

We interpret this provision as an obligation to segregate assets on-chain, off-chain, and legally (legally in the sense of establishing, for instance, a bailment with the clients’ crypto-assets). In this section, we discuss on-chain and off-chain commingling or segregation, while we discuss the legal treatment in insolvency in section IV.4.a.

62

On omnibus wallets regarding crypto-assets, see International Organization of Securities Commissions, ‘Final Report, Policy Recommendations for Crypto and Digital Asset Markets’ (November 2023) <https://www.iosco.org/library/pubdocs/pdf/IOSCOPD747.pdf> accessed 13 December 2024.

63

The legal nature of rights in crypto-assets is subject to an intense debate, may have the legal nature of proprietary right (probably ab indiviso-undivided property) or a contractual claim. See Jannik Woxholth and others, ‘Competing Claims to Crypto-assets’ (2023) 23 Uniform Law Review 7; Ignacio Tirado and Louise Gullifer, ‘Proprietary Rights and Digital Assets: A “Modest Proposal” from a Transnational Law Perspective’ (2024) 84 Law and Contemporary Problems (in press). On the property theories of NFTs, see Juliet M Moringiello and Christopher K Odinet, ‘The Property Law of Tokens’ (2022) 74 Florida Law Review 607, 647–57 <https://scholarship.law.tamu.edu/facscholar/2006> accessed 13 December 2024 and on the clients’ rights to the reserve assets of stablecoins, see Kara J Bruce, Christopher K Odinet and Andrea Tosato, ‘The Private Law of Stablecoins’ (2022) 54 Arizona State Law Journal 1073, 1123ff <https://scholarship.law.tamu.edu/facscholar/1999> accessed 13 December 2024.

64

For the implications of inaccurate books and records in case of crypto exchanges, see Levitin (n 10) 51–52.

65

ISDA, ‘Navigating Bankruptcy in Digital Asset Markets’ (2023) 16–17 <https://www.isda.org/a/CrLgE/Navigating-Bankruptcy-in-Digital-Asset-Markets-Digital-Asset-Intermediaries-and-Customer-Asset-Protection.pdf> accessed 13 December 2024.

66

Omnibus accounts commingle clients’ assets, while collective accounts commingle clients’ assets with the custodian’s own assets.

67

For instance, the keys associated with client A and its wallet address A may need to be stored in a USB stick different from the keys associated with client B and its wallet address B. Additionally, the keys associated with clients’ omnibus wallet address X may need to be stored in a different device from the keys associated with the custodian’s own assets wallet address Y.

68

We identified two cases where our sample service providers deny their wallet provision and or custodial function and merely mention the appointment of a third party for the safekeeping of their clients’ crypto assets.

69

One of those mentions that the ‘assets are not segregated’.

70

These include one which mentions that the ‘assets are not segregated’.

71

We also note that segregation by means of accounting methods is mentioned in three distinct contexts: (a) segregation from the custodian’s own assets in 11 sample T&Cs, (b) segregation from other clients’ assets in three sample T&Cs and (c) in three sample T&Cs we were not able to identify from whom the assets are segregated.

72

For details see Tirado and Gullifer (n 63); Woxholth and others (n 63) 7–19; Kara J Bruce and Christopher K Odinet and Andrea Tosato, ‘Stablecoins in Bankruptcy’ (2022) 42 Bankruptcy Law Letter 10.

73

The misconceptions and misleading practices concerning the omnibus or segregated wallets can lead to legal uncertainty or obstacles in the identification and recovery of the assets even for clients retaining proprietary rights.

74

Inter alia ISDA (n 65), 16; Louise Gullifer, Henry Chong and Hin Liu, ‘Client-Intermediary Relations in the Crypto-Asset World’ in Paul S Davies and Tan Cheng-Han, Intermediaries in the Commercial World (Hart 2022) 213–34; Yesha Yadav and Robert J Stark, ‘The Bankruptcy Court as Crypto Market Regulator’ (2024) 96 Southern California Law Review 1479; Justin Bernbrock, Jennifer Nassiri and Pouneh Almasi, ‘Ownership Issues in Crypto Cases’ (2023) 42 American Bankruptcy Institute Journal; Eva Micheler, ‘Custody Chains and Asset Values: Why Crypto-Securities are Worth Contemplating’ (2015) 7 Cambridge Law Journal 505.

75

For the additional risks of staged wallets, ie when the service provider appoints a third-party safekeeping the clients’ assets see Levitin (n 10) 58–60.

76

See Woxholth and others (n 63) 7–19. For an argument that streamlining of crypto insolvency laws is a priority, see Zetzsche and others (n 51) 91.

77

‘… your rights in relation to any On-Exchange Assets are limited to a contractual obligation for Crypto.com to provide an equivalent amount and type of On-Exchange Assets …’, Crypto.com, ‘Exchange Terms and Conditions’ (Crypto.com>, version 15 January 2024) <https://crypto.com/exchange/document/tnc>.

78

Gullifer and others (n 74).

79

For the fund depositary’s right of reuse, see AIFMD, article 21(8); UCITSD, article 22(5) and their respective delegated acts. For the MiFID custodian’s right of reuse, see MiFID, article 16 and the respective delegated acts.

80

See Zetzsche, Sinnig and Nikolakopoulou (n 6) 208–10.

81

MiCA, article 75(8).

82

The service providers explicitly cap any liability that may arise, up to the total amount of the fees that the client has paid to them, usually in the last 12 months before the incident.

83

This confirms the earlier finding by Awrey and van Zwieten (n 2) 41 (stating that they ‘faced significant obstacles in identifying even basic legal information for many [Shadow Payment Providers], let alone the more granular corporate, contractual, and other information needed to measure current levels of consumer protection’).

84

See Awrey and van Zwieten (n 2) 31 (applying a much broader concept of ‘shadow payment providers’, which besides centralized cryptoexchanges also includes mobile money service providers, remittance services, and proprietary payment service providers).

85

See Fidelity Digital Assets <https://www.fidelitydigitalassets.com/trading-custody> accessed 13 December 2024.

86

For retail market platforms, see the seminal article by Lina M Khan, ‘Amazon’s Antitrust Paradox’ (2017) 126 Yale Law Journal 710; for an application to financial services see Dirk Zetzsche and others, ‘Digital Finance Platforms: Toward a New Regulatory Paradigm’ (2020) 23 University of Pennsylvania Journal of Business Law 273.

87

See also Awrey and van Zwieten (n 2) 41 (stating the same conclusion while applying a much broader concept of shadow payment providers).

88

See Awrey and van Zwieten (n 2) 31–32.

89

SeeTim Copeland, ‘FTX claims rise above 100 cents on the dollar on back of bankruptcy plan’ (The Block, 8 May 2024) <FTX claims rise above 100 cents on the dollar on back of bankruptcy plan | The Block> accessed 13 December 2024.

90

See for an account of possible regulatory options for ‘shadow payment providers’, Awrey and van Zwieten (n 2); on ‘digital finance platforms’ see Zetzsche and others (n 86) 324–37.

Author notes

Dirk Zetzsche, Professor of Law, ADA Chair in Financial Law (Inclusive Finance), Department of Law, Faculty of Law, Economics and Finance, University of Luxembourg, 4, rue Alphonse Weicker, L-2721 Luxembourg; Co-Principal Investigator, FinTech National Centre of Excellence, Luxembourg. Tel: + 352 4666445092; Email: [email protected].

Areti Nikolakopoulou, Doctoral Researcher, Department of Law, Faculty of Law, Economics and Finance, University of Luxembourg, 4, rue Alphonse Weicker, L-2721, Luxembourg. Email: [email protected]. The authors are grateful for comments from participants and reviewers at the GLEA conference, as well as conferences held at the Universities of Vienna and Luxembourg. We thank Jenna Hensen, Diletta Blangero, Anida Mumdzic and Charles Koukabi for their research assistance. Further thanks are extended to Dan Awrey, Julia Sinnig, and Jannik Woxholth for their feedback on earlier drafts, and to Maurits van Ek for comments on our initial database. All remaining errors are our own.

Disclosure Statement: Neither Professor Dirk Zetzsche nor Areti Nikolakopoulou have provided advice to crypto custody providers. This research was funded in whole, or in part, by the Luxembourg National Research Fund (FNR), grant reference NCER22/IS/16570468/NCER-FT.

This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.