A wallet provider introducing account abstraction – programmable smart-contract wallets that bundle transactions, sponsor gas and rotate keys – discovers that the legal characterisation turns not on whether the architecture is "non-custodial" but on who holds recovery authority and whether the service resembles custody or payment facilitation under the applicable regime. In the European Union under MiCA, in Singapore under the Payment Services Act, and across the VASP frameworks in the UAE, Hong Kong and Switzerland, the question is the same: does the provider control, safeguard or intermediate the transfer of crypto-assets on behalf of users, and if so, does that control trigger a licence requirement? This analysis maps where account-abstraction models – including social recovery, session keys, bundled operations and paymaster gas sponsorship – cross the regulatory perimeter, and how providers structure to stay outside it or to secure the correct authorisation when they cannot.
The topic sits at the edge of two moving boundaries. First, the custody perimeter: when does holding a key fragment, a recovery shard or a meta-transaction signing right become safeguarding for regulatory purposes? Second, the payment-services perimeter: when does batching user operations or sponsoring gas fees make the provider an intermediary in a funds transfer? Regulators in the leading hubs now read these models through the lens of economic substance – who bears loss if the asset disappears, who can block a withdrawal, who sees the transaction before execution – rather than through labels such as "smart contract" or "self-custodial." The analysis that follows walks through the structural features of account abstraction (ERC-4337 and similar standards), compares how five flagship regimes classify them, identifies the design choices that shift a provider across the line, and sets out the decision matrix by operator profile and target market.
What is account abstraction, and why does it matter for regulation?
Account abstraction replaces the traditional externally-owned account – a private key controlling an on-chain address – with a smart-contract wallet. The user's "account" is itself a contract that can enforce rules: multi-signature approvals, spending limits, session keys with bounded permissions, social recovery (a set of guardians who can rotate keys if the original is lost), and transaction batching (bundling multiple operations into one on-chain call). The wallet contract can also delegate gas payment to a third party – a "paymaster" – so that a user transacts without holding the native token of the network. The architecture is designed to improve user experience: no seed-phrase custody anxiety, no gas-token hurdle for onboarding, programmable security policies instead of a single point of failure.
For a regulator, the question is this: if the provider operates the paymaster, holds a recovery key shard, or runs the bundler infrastructure that assembles and submits user operations to the blockchain, has that provider assumed a role that looks like custody, like payment intermediation, or like both? The answer depends on jurisdiction, but the analytical method is converging. MiCA, the EU's Markets in Crypto-Assets Regulation, defines custody of crypto-assets on behalf of clients as a regulated CASP activity; VARA in Dubai and the FSRA in Abu Dhabi's ADGM each define custody and transfer services as licensable activities; MAS in Singapore treats safeguarding and transmission of digital payment tokens as activities requiring a payment-institution licence; the SFC in Hong Kong and FINMA in Switzerland apply similar economic-substance tests. The label "non-custodial wallet" does not answer the question if, functionally, the provider can freeze funds, recover assets or see and alter transaction content before execution.
In our cross-border practice, we see three operator profiles attempting account-abstraction builds. Profile A: a custodial exchange or fund administrator adding programmable wallets as a user-experience feature – already licensed, the question is whether the new feature fits within the existing authorisation or requires a variation. Profile B: a wallet-as-a-service provider targeting enterprises (payroll in stablecoins, treasury automation) – currently unregulated, seeking to stay outside the perimeter or to secure a narrow licence that does not carry the full custody compliance load. Profile C: a consumer wallet startup using social recovery and gas abstraction – often structured as "fully decentralised" but functionally holding recovery authority or bundler control, and facing licensing questions in every market where it onboards users. The matrix below this section addresses each.
The custody perimeter: key control and recovery authority
Custody, across the flagship digital-asset regimes, is not defined by whether an entity calls itself a custodian but by whether it holds, controls or safeguards a client's crypto-assets. The test turns on three factual questions: (1) can the provider unilaterally move the asset? (2) can the provider block a client withdrawal? (3) does the provider hold information – keys, shards, recovery credentials – that would allow it to reconstitute control if the client's access is lost? If the answer to any of these is yes, most regulators will characterise the arrangement as custody or as a functional equivalent requiring similar authorisation.
Account abstraction complicates the analysis because control is distributed. A 2-of-3 multi-signature wallet where the provider holds one key and the user holds two does not, on its face, give the provider unilateral control – but if the provider also offers a social-recovery service and holds a guardian key, the combined position may be enough to trigger the custody perimeter. A session-key model – where a user delegates limited signing authority to a device or application for a set time or transaction count – looks like the user retaining control, but if the provider's infrastructure can revoke or alter those session keys, regulators may view the provider as an intermediary with safeguarding duties. A gas-sponsorship paymaster that merely pays transaction fees and cannot see or alter the user's operation is less likely to be classed as custodial, but a paymaster with discretion to approve, queue or reject operations begins to resemble a gatekeeper – and gatekeepers are regulated.
MiCA's definition, read through ESMA's technical standards (which are under consultation as this analysis is written), will likely hinge on whether the CASP has "exclusive control" or "joint control enabling unilateral action." A provider holding one shard in a threshold scheme where it cannot act alone may still be caught if the scheme is marketed as offering "secure custody" or if the provider's terms give it override rights in a dispute. VARA's custody definition in Dubai includes "holding or controlling Virtual Assets on behalf of another Person," interpreted broadly; a wallet provider that can freeze a user's smart-contract account through an admin function falls within that scope. In Singapore, MAS guidance on DPT services treats "safeguarding" as including arrangements where the service provider can reconstitute access – social recovery falls here if the provider holds enough shards.
The design choices that keep a provider outside the custody perimeter are these: (1) the user holds all key material necessary to move assets, and the provider's role is purely infrastructural (hosting the contract code, relaying signed transactions); (2) any recovery mechanism is entirely user-controlled (the user nominates guardians from their own contacts, and the provider has no override); (3) the provider cannot see private transaction details before execution, and cannot block or alter them. Providers that can meet these three conditions – and can demonstrate them in code and operational practice – generally remain outside the regulated perimeter, but they sacrifice two commercial features: the ability to offer custodial insurance (which requires the provider to be the insured party with control) and the ability to comply with a court order to freeze assets (which institutional clients often require).
Payment services and transfer: when the wallet intermediates
The second perimeter is payment services or funds transfer. If the account-abstraction provider does not just hold assets but also routes, batches or approves their movement, it may be classified as a payment-services provider, a money transmitter (in the United States) or a VASP conducting transfer or settlement services (under VARA, MiCA or the BVI/Cayman VASP Acts). The test here is whether the provider stands between payer and payee in a way that gives it discretion, timing control or visibility over the transaction.
A bundler – the off-chain service that collects user operations, simulates them and submits them on-chain as a batch – is central to ERC-4337 account abstraction. If the bundler is run by the wallet provider and the provider decides which operations to include, in what order and at what gas price, the provider is intermediating. If the bundler is decentralised (multiple independent operators competing to submit bundles) and the user can choose or rotate bundlers freely, the intermediation risk is lower – but the provider must demonstrate that decentralisation in practice, not just in documentation. ESMA's draft technical standards for MiCA are expected to address whether operating a bundler constitutes "execution of orders" or "arranging transactions," both of which are regulated CASP activities if conducted on behalf of clients. Operators in that position will need either a full CASP authorisation or a clear exemption (e.g. if the bundler acts as a pure infrastructure service with no client relationship).
A paymaster that sponsors gas introduces a second transfer: the user's intended operation (say, a stablecoin payment) and the paymaster's separate gas payment. If the paymaster is controlled by the wallet provider and the provider's terms allow it to reject sponsorship based on transaction content, compliance flags or user status, the provider is exercising discretion in the payment flow – a characteristic that several regulators associate with money transmission or payment intermediation. If, by contrast, the paymaster sponsorship is automatic, non-discretionary and blind to transaction content (the provider simply covers gas for any valid operation from its users), the payment-services risk is smaller, though not zero; MAS in Singapore has indicated that even non-discretionary facilitation can trigger DPT service classification if it is systematic and customer-facing.
The structural answers depend on the provider's business model. A wallet provider that charges enterprises a SaaS fee for paymaster sponsorship (a flat monthly rate, not per-transaction) and cannot see or control the underlying payments is closer to an infrastructure vendor than a payment intermediary. A wallet provider that takes a fee on every transaction, holds a float to cover gas and can suspend service to individual users based on risk scoring is closer to a regulated payments business. In a recent structuring matter, an enterprise wallet platform separated its bundler operation (structured as an independent network participant with no customer relationship) from its commercial wallet service (which had customer contracts but no bundling discretion), precisely to keep the payment-services question off the table while the provider pursued a narrow e-money or stablecoin-issuance licence in an EU member state.
Session keys and delegation: the thin line between UX and intermediation
Session keys – time-limited, scope-limited signing keys that a user creates within their smart-contract wallet to authorise specific actions without exposing the master key – are marketed as a user-experience improvement and a security feature. A user might generate a session key that allows a gaming application to spend up to 100 USDC on in-game purchases over the next seven days, without requiring the user to sign every micro-transaction. The master key remains with the user; the session key lives on the user's device or in the application's enclave; the wallet contract enforces the spending cap and expiry on-chain.
The regulatory risk arises if the wallet provider, not the user, controls session-key creation, scope or revocation. If the provider's backend can unilaterally generate a session key, that looks like the provider holding signing authority – functionally equivalent to holding a private key, and therefore custody. If the provider can revoke or extend a session key outside the user's control, the provider is a gatekeeper. If session keys are user-created and user-managed, enforced entirely by the on-chain wallet contract with no provider discretion, the activity remains with the user – but the provider must then ensure that its terms, its code and its operational logs all reflect that the provider never touches session-key material and cannot override user choices.
Where session keys interact with gas sponsorship, the picture thickens. Suppose a gaming platform uses a session key to authorise in-game token transfers and a paymaster to cover gas. The user signs nothing at transaction time. From the user's perspective, the platform is "handling" the transaction, even though the on-chain execution is the user's wallet contract calling a game contract. If a regulator views the combined session-key-plus-paymaster setup as the platform acting on the user's behalf – with the user having delegated transaction authority in a way that removes real-time control – the platform may be conducting a regulated activity (custody, payment services or both). The platform's defence is that the session key is a user-created credential, not platform-held, and that the paymaster is a separate, non-custodial service. That defence holds if the architecture genuinely separates the two and if the provider never holds a master or recovery key. If the provider also offers a backup-and-recovery service and holds a guardian shard, the cumulative control tips toward regulation.
Social recovery: the regulatory lightning rod
Social recovery – the feature that allows a user to recover their wallet if they lose their key by obtaining approval from a set of pre-nominated guardians – is both the strongest user-experience selling point of account abstraction and the feature most likely to place the provider inside the custody perimeter. The mechanic: the user nominates three to five guardians (friends, family, other devices, or a professional recovery service); if the user loses access, a threshold of guardians (say, three of five) can approve a key rotation or account reset, typically by signing a recovery transaction that the wallet contract verifies on-chain.
If the wallet provider is one of the guardians – or if the provider operates a default guardian service that users are encouraged to include – the provider holds recovery authority. A regulator reading MiCA's custody definition, VARA's "control" language or MAS's "safeguarding" concept will conclude that the provider can reconstitute access to the user's assets, which is the economic substance of custody. The provider's argument that it is only one of several guardians, and cannot act unilaterally, does not remove the issue; joint control with the ability to facilitate recovery is still a form of control. Custodians in traditional finance are often joint signatories or operate under dual-control requirements – that does not make the activity non-custodial.
The cleaner structure: the provider offers social-recovery functionality as a feature of the wallet contract, but does not itself act as a guardian and has no mechanism to override or participate in a user's recovery. The user selects guardians from their own contacts (email addresses, phone numbers, other wallets); the provider's code enables the on-chain verification of guardian signatures, but the provider never holds a guardian key and cannot initiate or approve a recovery. This keeps the provider in the role of a software vendor, not a custodian. The trade-off: users bear full responsibility for guardian selection, and if a user loses access and has chosen guardians poorly (or if guardians collude), the provider cannot help – and that loss of "customer support as a safety net" is often a commercial deal-breaker for consumer products. Providers targeting institutional clients find this trade-off easier; enterprises typically prefer to hold their own guardian keys or to split them across internal signatories and an independent professional custodian, rather than rely on the wallet vendor.
Jurisdiction by jurisdiction: how the perimeter is drawn
MiCA (European Union) classifies custody of crypto-assets on behalf of clients as a CASP activity requiring authorisation. The drafting captures "safeguarding" of crypto-assets and "control of means of access" (private keys or equivalent). A wallet provider that holds key shards enabling recovery, or that operates infrastructure with the practical ability to block transactions, will likely need a CASP licence for the custody leg. If the provider also bundles or routes transactions on behalf of users, the "execution of orders" or "placing of orders" CASP activities may apply. Passporting across the EU is available once authorised in one member state, but the application process – particularly the own-funds requirement and the operational-resilience standards – is demanding. Providers serving European users at scale typically apply in a member state with digital-asset supervisory experience (Lithuania, Malta, France, the Netherlands) or structure as a pure software provider with no key-holding, no recovery service and no bundler control, avoiding the CASP perimeter entirely.
Singapore (MAS) distinguishes between a digital payment token service and a pure software provision. Safeguarding or transmitting DPTs on behalf of customers triggers licensing. MAS has indicated in past guidance that holding key fragments sufficient to reconstitute access is safeguarding, even if the customer holds other fragments. A wallet provider using threshold cryptography (e.g. 2-of-3, where the provider holds one shard and the user holds two) is likely caught unless the provider can demonstrate it has no ability to act without user participation and no recovery override. Bundler operation, if it involves discretion over transaction inclusion or ordering, may be viewed as transmission or facilitation. Providers targeting Singapore users generally secure a standard or major payment institution licence (depending on transaction volume) or, if they wish to remain unlicensed, ensure that all cryptographic material is client-held and that the provider's role is limited to hosting open-source contract code.
United Arab Emirates – Dubai (VARA) treats custody and transfer/settlement as separate licensable activities. Custody is broadly defined: holding or controlling virtual assets on behalf of another person. A wallet provider that can freeze a user's smart-contract account, recover assets via a guardian role or route transactions through provider-controlled infrastructure is conducting a regulated activity. VARA's operational rules (the VARA rulebooks for each activity) require detailed policies on safeguarding, insurance, segregation and incident response. A provider that crosses the line must apply for the relevant VARA licence, a process that includes a formal application, a detailed compliance manual, background checks on controllers and senior management, and ongoing supervision. Operators we advise in the UAE typically choose between a full custody licence (if they are building a custodial product for institutions and want the credibility that a VARA licence confers) or a non-custodial software-vendor structure (if they are building for retail and cannot carry the compliance cost).
Hong Kong (SFC) licenses virtual-asset trading platforms (VATPs) and has signalled that wallet services with custodial elements – particularly if marketed to Hong Kong retail users – will face licensing expectations. The SFC applies a substance-over-form test: if the service provider can control, recover or intermediate the user's assets, the arrangement is custodial or quasi-custodial, and the provider should either hold a VATP licence (if it also offers trading) or seek an appropriate authorization under the securities or anti-money-laundering regimes. Account-abstraction providers serving Hong Kong users typically structure outside Hong Kong (common destinations: Singapore, BVI, Cayman) and treat Hong Kong as a restricted jurisdiction unless and until they are prepared to apply for a licence.
Switzerland (FINMA) does not have a standalone "wallet licence," but a custodial wallet service is a financial service under Swiss law, particularly if the provider holds assets or provides safeguarding that resembles banking activity. Providers may pursue a fintech licence (lower capital, capped deposit/wallet volume) or affiliate with a self-regulatory organization (SRO) for anti-money-laundering supervision, depending on scale. A non-custodial wallet that is genuinely user-controlled and cannot recover or block assets typically remains outside FINMA's direct perimeter, but the provider must still comply with AML/KYC rules if it onboards users systematically. FINMA's approach is principles-based: if the economic substance is custody, it will be regulated as custody, regardless of the technical architecture.
The compliance overlay: AML, Travel Rule, sanctions
Even if a wallet provider structures to stay outside the custody and payment-services perimeters, it faces a separate compliance question: is it a virtual asset service provider (VASP) for anti-money-laundering purposes, and does the Travel Rule apply? The FATF definition of a VASP includes entities that "conduct one or more" of several activities as a business on behalf of another person, including transfer, safeguarding and exchange. A non-custodial wallet vendor – one that truly holds no keys and cannot control user assets – is not, under the FATF definition, a VASP. But if the vendor holds recovery shards, operates the bundler with discretion, or provides paymaster services that allow it to see and potentially block transactions, it crosses into VASP territory.
The Travel Rule (FATF Recommendation 16) requires VASPs to pass originator and beneficiary information when conducting a transfer above a de-minimis threshold (typically USD 1,000, though the threshold varies by jurisdiction). If a wallet provider is classed as a VASP and if user-to-user transactions through the wallet constitute "transfers" that the provider intermediates (because the provider's bundler submits them, or because the paymaster's involvement gives the provider visibility and control), the provider must collect and transmit Travel Rule data – a heavy operational and technical burden. In practice, most consumer wallet providers avoid this outcome by ensuring that they never intermediate: the user signs, the user broadcasts (or the wallet relays a fully-signed transaction without discretion), and the provider cannot see or alter transaction content. Providers that cannot avoid intermediation – typically enterprise wallet-as-a-service platforms that bundle transactions for efficiency – accept the VASP classification and build Travel Rule pipes (often using a third-party Travel Rule solution such as those offered by Notabene, Sygna or TRP Lab, though we name these purely as examples of the category, not as endorsements).
Sanctions screening introduces a parallel question. If the wallet provider has the ability to block a transaction – either through bundler discretion or through a smart-contract admin function – does it have a legal obligation to screen against OFAC or EU sanctions lists, and is it liable if a sanctioned actor uses the service? The answer depends on whether the provider is a US person or has a US nexus (in which case OFAC's "facilitation" prohibition applies), whether the provider is an EU operator subject to EU sanctions regulations, and whether the provider's terms of service create a contractual duty to comply with sanctions. Providers that genuinely cannot block transactions – because the wallet contract is immutable and the provider has no admin key – have a strong argument that they lack the technical capability to comply with a blocking obligation, though they remain subject to "know your customer" obligations if they are a VASP. Providers that retain admin control accept that they are expected to maintain a sanctions-screening programme, typically by screening wallet addresses at onboarding (using a blockchain-intelligence vendor) and by monitoring for OFAC SDN list additions.
Structuring the account-abstraction provider: the entity stack and liability splits
For a wallet provider that has decided where it sits on the custody and payment-services spectrum, the next question is entity structure. The provider typically splits functions across entities to isolate regulated activities, limit liability and optimise tax. A common stack: (1) a software-development company (often in a tech-neutral, low-tax jurisdiction such as the Cayman Islands or the BVI) that owns the intellectual property and licenses the wallet code; (2) a regulated entity (in an EU member state if serving Europe, in Singapore if serving Asia, in Dubai or Abu Dhabi if serving MENA and Asia, or in multiple jurisdictions if the provider wants full global coverage) that holds the CASP or payment-institution licence and onboards users in regulated markets; (3) a bundler operation structured as a separate network participant or as an independent subsidiary with no customer-facing role, to keep payment-services regulation away from the software business; (4) a foundation or DAO (if the protocol has a governance token) that holds certain admin keys or protocol-upgrade rights, separating the commercial business from the decentralised governance layer.
The liability concern is this: if a user loses funds because of a smart-contract bug, a bundler failure or a paymaster error, who is liable, and under what theory? If the wallet provider holds no keys and cannot recover assets, it will argue that it is a software vendor, not a fiduciary, and that liability is limited to software defects (which the provider will attempt to exclude or cap in its terms). If the provider holds recovery keys, operates the bundler or sponsors gas with discretion, a court or regulator is more likely to find a duty-of-care relationship – particularly if the service is marketed as "secure" or "backed by professional custodians." Providers in that position typically obtain professional-indemnity and errors-and-omissions insurance (at meaningful limits, which for a custody-adjacent service often start in the low eight figures), and they build incident-response and customer-claim procedures that mirror those of a licensed financial institution.
In a structuring matter during the northern-hemisphere autumn, a wallet provider offering session keys and social recovery to enterprises separated the custody and non-custody functions. The custodial leg – holding one shard in a 2-of-3 guardian setup – was moved into a licensed VASP entity in the BVI (chosen for the combination of a clear VASP Act, a faster licensing timeline than an EU CASP application, and acceptance by banking partners). The non-custodial wallet-as-a-service product – where enterprises held all key material and the provider offered only infrastructure – remained in a Cayman software company, unlicensed. The split allowed the provider to serve two markets: institutional clients who wanted a licensed custodian as one guardian (and who were prepared to pay custody fees), and self-custodial enterprises who preferred lower cost and full control. The same provider then applied for a MiCA CASP licence in Lithuania to serve EU clients under the passporting regime, a process that is ongoing as this analysis is finalized.
What about "full decentralisation" – does it remove the provider entirely?
Some wallet projects adopt the position that if the account-abstraction protocol is fully open-source, if the smart-contract wallet code is immutable, if bundlers are run by independent third parties, and if no single entity controls the system, then there is no "provider" to regulate. This framing – common in the decentralized-finance (DeFi) community – relies on the idea that code is speech, that deploying an immutable contract is not offering a service, and that regulatory frameworks are designed for intermediaries, not for autonomous systems. The legal counterargument, advanced by regulators in the United States (the SEC and CFTC in various enforcement actions), in the EU (under MiCA's provisions on "arranging or advising"), and in Singapore (MAS guidance on DeFi), is that if a team developed the protocol, marketed it, profits from it (through a token or a fee split), or retains any upgrade or admin rights, that team is the functional operator and can be held accountable as a service provider, regardless of the protocol's technical decentralisation.
The practical reality is that "full decentralisation" is a spectrum. A protocol where the founding team has renounced all admin keys, where the code has been audited and is immutable, where the user interface is hosted on IPFS and can be accessed permissionlessly, and where the founding team has wound down and holds no tokens is at one end – genuinely decentralised, and difficult for a regulator to bring within a service-provider framework. A protocol where the founding team controls a multi-signature wallet that can upgrade contracts, where the team runs the primary bundler and website, where the team's entity collects paymaster fees, and where the team markets the wallet to consumers under a brand is at the other end – functionally a centralized service with a decentralized-technology back end, and clearly within the perimeter. Most projects in the market today sit between these poles: the code is open-source, but the team provides a hosted front end and takes a fee; the smart contracts are upgradeable via a DAO vote, but the founding team controls enough governance tokens to pass proposals; the bundler network is "permissionless," but in practice the team's infrastructure processes the majority of transactions.
Our view, reflected in the structuring work we do for account-abstraction and DeFi clients, is that decentralisation is a risk-mitigation feature, not a get-out-of-regulation card. A provider that achieves genuine decentralisation – and can demonstrate it through on-chain data, governance logs and a credible narrative of team dissolution or pivot to pure governance – reduces its regulatory risk significantly, particularly in jurisdictions that take a principles-based, substance-over-form approach. But the path to that level of decentralisation is multi-year, requires giving up product control and revenue optionality, and is incompatible with venture-capital return expectations in most cases. Providers that want to remain accountable (and investable) typically accept that they will be regulated and build accordingly.
Decision matrix: which structure for which operator?
Profile A: Licensed custodian adding account-abstraction features. You are a regulated exchange, a fund administrator or a qualified custodian, already licensed under MiCA, MAS, VARA or an equivalent regime. You want to offer clients programmable wallets – multi-signature, session keys, gas abstraction – as a user-experience improvement and a differentiator against competitors. Key question: does the new feature fit within your existing license scope, or does it require a variation or a new authorisation? In most cases, if you already hold custody authorisation, adding smart-contract wallet capabilities is an operational enhancement, not a new regulated activity, provided that you remain the entity holding keys (even if those keys are now split via threshold cryptography or delegated via session keys under your control). The compliance lift is updating your operational procedures, your incident-response playbook and your insurance policy to cover smart-contract risk. If your paymaster sponsors gas for client transactions, ensure that your AML transaction-monitoring system can attribute those operations to the underlying client – regulators will expect you to monitor the substance of the transfer, not just the gas payment. Timeline: three to six months to integrate account-abstraction into your existing custody stack, assuming that your technology vendor or in-house team has the capability. No new licence if you are already authorised for custody. Risk: smart-contract bugs creating client losses; ensure that your contracts are audited by a reputable firm (examples of firms in this space include Trail of Bits, OpenZeppelin, ConsenSys Diligence – named as category references, not recommendations) and that your insurance underwriter is comfortable with the added exposure.
Profile B: Wallet-as-a-service for enterprises, currently unlicensed. You provide wallet infrastructure to companies for payroll, treasury or payment-application use cases. You do not hold client keys today, but you are considering adding social recovery or paymaster services to improve your product's competitiveness. Key question: will those features push you into the custody or payment-services perimeter, and in which target markets? If you add social recovery where you hold a guardian key, you become a custodian in most jurisdictions and will need a licence (CASP in the EU, payment institution in Singapore, VASP in the UAE or BVI, depending on where your clients sit). If you add a paymaster where you have discretion over transaction inclusion, you may be a payment-services provider. Our recommendation: if your clients are enterprises and your deal sizes justify it, pursue a narrow VASP or payment-institution licence in one hub jurisdiction – commonly the BVI (faster, clearer VASP Act, accepting of non-resident applicants, twelve to sixteen weeks to authorisation) or Lithuania (MiCA transition in progress, EU passporting, reasonable application timeline). Structure your bundler as a separate network operation or outsource it to a third-party infrastructure provider to keep the payment-intermediation question off your licence application. If your clients are price-sensitive and your volumes do not support the compliance cost of a licence, build the product without holding any keys and without bundler discretion – offer social recovery as a user-controlled feature (the user selects guardians; you provide the contract logic) and offer paymaster as a flat-fee SaaS service (you cover gas blindly; you do not see transaction content). Timeline: licensing path, nine to eighteen months; non-custodial build, six to twelve months.
Profile C: Consumer wallet with decentralisation narrative. You are a startup targeting retail users, using account abstraction to remove seed-phrase anxiety and to onboard users without requiring them to hold ETH or another gas token. Your pitch is "decentralised and user-controlled," but your backend runs the bundler and holds one key in a 2-of-3 recovery setup. You have venture funding and a token launch planned. Key question: can you credibly maintain that you are not a regulated custodian or payment-services provider, or do you need to shift your structure before a regulator (or a cease-and-desist letter) forces the issue? Our assessment: you are on the line, and which side you fall on depends on your exact technical design and your target markets. If you are marketing in the EU, Singapore, Dubai, Hong Kong or the United Kingdom, expect that a regulator will eventually ask whether you need authorisation – particularly if you gain traction and press coverage. The path forward depends on your risk appetite. High risk tolerance: proceed as-is, structure your entity in a jurisdiction with no clear wallet-provider regime (certain offshore centres remain in that category, though the number is shrinking), block users from high-enforcement jurisdictions (US, EU, SG, HK), and be prepared to pivot to full decentralisation or to licensing if required. Medium risk tolerance: remove your entity from the recovery process (make social recovery purely peer-to-peer, so you are a software vendor only), decentralise the bundler (accept that transaction times may be less reliable), and document your non-custodial posture clearly in your terms, your marketing and your published architecture diagrams. Low risk tolerance: treat yourself as a regulated entity from day one, incorporate in a licensing-friendly jurisdiction, apply for the relevant authorisation (typically a payment-institution or VASP licence in a Tier 2 hub with reasonable timelines and costs), and build your compliance programme concurrently with your product. The low-risk path is slower and burns more capital upfront, but it allows you to market openly in major economies and to work with institutional partners (exchanges, fiat on-ramps) who require their wallet-integration counterparties to be licensed.
Common mistakes and how to avoid them
Mistake one: labeling the product "non-custodial" in marketing while retaining technical control. Regulators do not accept labels; they read the code, the terms of service and the operational setup. If your smart contract has an admin key that your entity controls, if your backend can trigger a recovery without user participation, or if your bundler can censor transactions, you are custodial or quasi-custodial in substance, and calling yourself non-custodial is a regulatory red flag. Fix: align your marketing with your technical reality, or change the technical reality to match the marketing.
Mistake two: assuming that operating across borders from an offshore entity removes regulatory risk. Digital-asset regulation increasingly follows users, not providers. If you onboard users in the EU, Singapore or Dubai, the relevant regulator will take the view that you are offering a service in that jurisdiction and must comply with local rules, regardless of where your entity is incorporated. Several enforcement actions against unlicensed exchanges and wallet providers have proceeded on exactly this theory. Fix: either secure authorisation in each major market where you have users, or implement robust geoblocking (IP blocking, KYC-based jurisdiction checks, terms-of-service prohibitions) and accept that you are limiting your addressable market.
Mistake three: building the product first and addressing legal afterwards. Account-abstraction design choices – who holds keys, how recovery works, how the bundler operates, whether the paymaster is discretionary – are not just technical decisions; they are legal characterisation decisions. If you build a product where you hold recovery authority and then try to obtain a licence, you will find that the compliance cost (capital requirements, insurance, operational policies) is higher than your funding round contemplated, and restructuring the product post-launch is expensive and often impossible without alienating your user base. Fix: involve legal counsel at the architecture and specification stage, before you write production code. Map the regulated perimeter for your target markets, decide which side of it you want to sit on, and design accordingly.
Mistake four: underestimating the AML and Travel Rule burden. Even if you avoid custody and payment-services licensing, you may still be a VASP for AML purposes if your bundler or paymaster gives you intermediation capability. The Travel Rule is particularly complex for wallet providers, because it requires collecting data about both the originator and the beneficiary, and in a peer-to-peer wallet context the provider may not have a relationship with the beneficiary. Providers that are surprised by this obligation post-launch often find themselves unable to comply without rebuilding their transaction flow. Fix: model your AML and Travel Rule obligations early, determine whether you can comply with your current architecture, and either redesign to avoid VASP classification or budget for the Travel Rule infrastructure (data collection, transmission to counterparty VASPs, record-keeping).
When to engage counsel and what to expect
The right time to engage legal counsel on account-abstraction structuring is during the product-specification phase – after you have a working prototype and a roadmap, but before you launch publicly and certainly before you take institutional investment that will expect a clear regulatory path. The questions counsel will ask: who holds keys, how does recovery work, how does the bundler operate, do you have discretion over transaction inclusion or content, where are your users, what is your entity setup, what are your funding and revenue expectations, and what is your risk appetite? The output is a regulatory characterisation – a memo that walks through the custody and payment-services perimeters in your target markets, identifies where your design crosses the line, maps the licensing options (with timelines, costs and strategic trade-offs), and proposes a structure that aligns with your business model and your investors' expectations.
For a straightforward non-custodial wallet (user holds all keys, provider is purely infrastructural), the engagement is short – often a fixed-scope opinion at a transparent fee, delivered in two to four weeks. For a complex enterprise wallet-as-a-service with paymaster and social recovery features, the work is staged: legal characterisation, then entity structure, then licence application (if required), then compliance-programme buildout, each phase priced and scoped separately. Providers in that category should budget for legal as a material line item – comparable to a security audit or a first-year insurance premium – and should expect the legal workstream to run concurrently with product development, not sequentially after launch.
The process above describes the standard path for a wallet provider meeting the issue for the first time. If a prior build crossed into the custody perimeter inadvertently, or if you launched without a licence and a regulator has now contacted you, the engagement is a triage: assess the current exposure (what the regulator will likely require, what the penalty risk is, what the timeline is), then map the quickest path to compliance (often a voluntary application combined with operational changes to reduce the scope of the conduct). In a recent recovery of this kind, a consumer wallet that had launched with a 2-of-3 recovery model (provider held one key) received an inquiry from a European NCA. We restructured the product to remove the provider key (converting to a user-peer social-recovery model where the provider facilitated the contract logic but held no key material), submitted a detailed response demonstrating the change, and the matter closed without penalty. The lesson: regulators in most hubs prefer to see voluntary compliance over enforcement, but the window to restructure is short once the inquiry arrives.
If your account-abstraction build crosses the perimeter and you need to map the licensing and structuring path, contact OBOLUS at info@oboluslaw.com. We scope each engagement transparently, and the first strategy call – under NDA – is without charge.
Related at OBOLUS
- DeFi, Tokenization & Smart-Contract Law – Regulatory structuring, liability allocation and compliance design for protocols and on-chain instruments.
- Tokenized Fund Share Classes: Structuring and Transfer Restrictions – Building on-chain share classes with regulatory transfer controls for institutional funds.
- Oracle and Data-Feed Liability in Australia – Managing third-party data-feed risk in automated on-chain systems under Australian regulatory settings.
FAQ
Can a decentralised protocol be regulated?
Yes, if a team developed, deployed, marketed or profits from the protocol and retains any control (upgrade keys, governance tokens, fee collection), that team can be held accountable as the functional operator under most digital-asset regimes. "Decentralised" describes the technology, not the legal relationship. Regulators in the EU, Singapore, the UAE and the United States have each taken enforcement or supervisory action against teams behind ostensibly decentralised protocols, applying a substance-over-form test. A protocol that is genuinely decentralised – no team control, no ongoing profit to the originators, immutable code – is harder to regulate, but such protocols are rare in practice. Most projects today retain enough control to be classified as service providers.
Who is liable when a smart contract fails?
Liability depends on the relationship between the user and the entity that deployed or operates the contract. If the contract is part of a commercial service (a wallet provider's account-abstraction contract, a custodian's multi-signature vault), the provider is typically liable for defects under a negligence or contract theory, subject to any limitation-of-liability clauses in the terms of service (which may or may not be enforceable, depending on jurisdiction and whether the failure caused consumer harm). If the contract is open-source and deployed by an anonymous team with no ongoing relationship to users, finding a liable party is difficult – users bear the risk. Insurance (protocol cover through providers such as Nexus Mutual or InsurAce, or traditional errors-and-omissions insurance held by the deploying entity) is often the only recovery route in a pure smart-contract failure, and even that is limited by policy exclusions.
What wrapper suits this structure?
The choice of legal entity for an account-abstraction provider depends on whether the provider holds keys (and therefore needs a regulated-entity wrapper) and where its users sit. A non-custodial software provider typically incorporates in a low-tax, business-friendly jurisdiction – the Cayman Islands, the British Virgin Islands, or Singapore – as a limited company or an exempted company. A custodial or payment-services provider needs a regulated entity in a jurisdiction that offers the relevant licence – an EU member state for MiCA CASP, Singapore for payment-institution authorisation, Dubai for VARA, the BVI or Cayman for VASP registration. Many providers use a holding-company structure: a Cayman parent (owning IP, holding investment), a regulated subsidiary (EU, Singapore or UAE, holding the licence and onboarding users in those markets), and potentially a separate entity for the bundler or protocol-governance functions. The structure is guided by tax efficiency, investor expectations, banking access and regulatory segmentation.
OBOLUS is an independent digital-asset law boutique acting only for businesses. We advise exchanges, custodians, token issuers and funds on licensing across more than seventy jurisdictions, on disputes and on-chain asset recovery across more than twenty-five forums, and on the tax, banking and compliance architecture that sits around them. Digital assets are the whole of our practice. Before you commit to an account-abstraction architecture, we pressure-test the regulatory perimeter and map the licensing path – or the design changes that keep you outside it. We work on transparent fixed-scope packages, and the first strategy call under NDA is without charge. To discuss your situation, contact info@oboluslaw.com.
By Roman Levitt, Technology & DeFi Counsel – Specialising in the legal classification of smart-contract protocols and the structuring of on-chain services across custody, payment and DeFi perimeters.
This publication is general information about the law and does not constitute legal advice. It is not a substitute for advice tailored to your circumstances. OBOLUS accepts no liability for action taken or not taken on the basis of this material. For advice on your situation, contact info@oboluslaw.com.