Functional Analysis of Crypto-Asset Service Providers in AML Frameworks: How to Qualify an Actor within a Non-Linear Architecture

The AML toolkit was built for a world of banks, custodians, and payment institutions linked in linear chains of intermediation. Crypto-asset systems, by contrast, are non-linear, protocol-driven, and functionally fragmented. This article argues that qualifying actors in such an environment requires a functional, dynamic, and architecture-aware methodology rather than simple transposition of legacy categories.

The regulatory architecture of anti-money laundering (AML) frameworks is historically anchored in a model of financial intermediation that predates the emergence of decentralised technologies. In this traditional framework, the identification and classification of actors is straightforward: banks, payment institutions, custodians, and money transmitters form a constellation of well-defined entities whose legal personhood and organisational structures make them suitable targets for AML obligations. The relationships among these actors are linear, hierarchical, and predictable.

The crypto-asset ecosystem, in contrast, unfolds within a radically different topology. Rather than a linear chain of intermediaries through which value passes, decentralised systems are characterised by non-linear networks, modular components, and fluid roles. Smart contracts, protocols, token issuers, validators, liquidity providers, and front-end operators form a constellation of functions that are neither clearly separable nor easily mapped onto the traditional categories of AML law. The FATF’s framework of Virtual Asset Service Providers (VASPs), although intended to capture actors within this new ecosystem, relies heavily on analogies with traditional intermediation. It therefore encounters structural limitations when confronted with the functional plurality and non-linear architecture of crypto-networks.

This article undertakes a functional analysis of crypto-asset actors for AML purposes. It argues that the qualification of actors within decentralised systems demands a methodological shift: doctrinal categories built for linear intermediation cannot simply be transposed onto an ecosystem structured around distributed and overlapping functions. Instead, regulators must adopt a dynamic, function-specific, and context-sensitive approach that recognises the fluidity of roles and the decentralised nature of control in blockchain-based systems.

I. The Traditional AML Model and Its Linear Functional Logic

Classical AML frameworks rest on a linear conception of financial operations. A user interacts with a financial intermediary; the intermediary processes the transaction; value flows from one account to another through controlled channels. The essential functional elements—custody, transfer, exchange, and settlement—are mediated by institutions entrusted with compliance duties. Each element of the transaction chain is associated with a corresponding regulated actor.

This model allows for clear attribution of obligations. Custodians conduct KYC, exchanges monitor trading activity, and payment institutions report suspicious transactions. The functions are decomposable and assignable to discrete entities. Importantly, the architecture presupposes that each transaction must transit through institutional chokepoints where compliance can be enforced.

The crypto-asset ecosystem disrupts this logic. Its architectural features undermine the linearity presupposed by traditional AML frameworks. Decentralised networks do not impose mandatory passage through intermediaries; functions can be executed programmatically; and the same economic operation may be performed simultaneously by multiple actors playing partial or overlapping roles. Understanding the functional nature of these roles is therefore essential.

II. The Non-Linear Topology of Crypto-Asset Systems: Functions without Institutions

Blockchain-based systems are typically designed around protocols rather than institutions. Their architecture is modular, meaning that custody, execution, and settlement may occur independently of any legal entity. Several features illustrate the non-linear nature of these systems.

First, execution and settlement are inseparable within decentralised systems. Transactions are validated by distributed consensus mechanisms, not by institutional operators. Validators perform settlement functions collectively, without assuming custody or exerting discretionary authority over the content of transactions. Their role is systemic rather than custodial.

Second, custody may be algorithmic rather than institutional. Users may retain control of their private keys, rendering “self-custody” the default mode rather than the exception. Smart contracts, in turn, may hold assets in an autonomous manner, without any human exercising factual control. This challenges the AML assumption that assets are always held by identifiable custodians.

Third, exchange and intermediation can occur through protocols rather than platforms. Decentralised exchanges (DEXs) allow the direct trading of tokens through automated market makers (AMMs). The execution of the trade does not depend on a central operator but on deterministic rules encoded in smart contracts. The front-end interface may provide access, but it is not essential to the functioning of the protocol.

Fourth, issuance and governance are frequently decentralised. Token issuers may disappear or relinquish control, allowing the protocol to evolve through decentralised governance decisions. The functional role of “issuer” may therefore dissipate over time.

These features make it difficult to map crypto-asset actors onto traditional AML categories. The functional logic of decentralised networks is non-linear, overlapping, and dynamic. Regulators cannot rely solely on the presence of identifiable institutions, as the architecture deliberately avoids centralised control.

III. FATF’s Functional Approach to VASPs: Doctrinal Ambition and Methodological Limits

Aware of the inadequacy of institutional analogies, the FATF introduced a functional definition of VASPs, seeking to classify actors based on the activities they conduct rather than their organisational structure. In theory, this functional approach should accommodate decentralised systems. In practice, however, the FATF’s methodology retains assumptions rooted in traditional financial regulation that limit its effectiveness.

The FATF identifies activities such as exchange, transfer, safekeeping, and administration of virtual assets. An actor is deemed a VASP if it “conducts” one of these activities “as a business.” Yet this approach presupposes that activities are performed by entities capable of understanding, controlling, and benefiting from them. It is designed for systems in which actors intentionally perform functions rather than situations in which functions emerge from the autonomous behaviour of protocols.

Moreover, the FATF struggles with the attribution of activities in decentralised systems. The notion of “conducting” an activity presupposes a degree of agency. But in decentralised networks, agency may be collective, distributed, or absent altogether. For example, liquidity providers in an AMM may not “conduct” exchange services; they merely supply liquidity. The exchange is performed by the algorithm, not by a human intermediary. Similarly, validators do not “conduct” transfer activities in the sense of executing orders; they merely verify transactions according to protocol rules.

The FATF attempts to salvage its framework by introducing the concept of “owners/operators,” suggesting that those with sufficient influence over a protocol may be responsible for VASP obligations. But influence is neither uniform nor easily measurable. Developers may have created a protocol but relinquished control; governance token holders may vote on certain changes but not control day-to-day operations; front-end providers may facilitate access without influencing outcomes. The FATF’s functional approach thus encounters conceptual limits when activities are not tied to identifiable agents.

IV. Functional Classification in Non-Linear Architectures: Toward a Methodological Reorientation

Given these difficulties, a functional analysis of actors in decentralised systems must depart from the assumptions of traditional AML frameworks. Two methodological principles emerge: first, functions must be analysed independently from institutional structures; second, attribution must account for the fluidity and multiplicity of roles within decentralised ecosystems.

1. Decomposing functions without presupposing institutions

In decentralised networks, functions—such as custody, transfer, exchange, issuance, and governance—exist independently of institutional operators. A functional analysis must therefore begin by identifying how these functions are instantiated technologically. For example, custody may be algorithmic (smart contract-based) or self-directed (private key control). Exchange may be executed by AMMs or through order-matching algorithms without discretionary intervention. Settlement is collective and consensus-based.

The functional nature of these operations must be assessed without presuming the existence of operators capable of carrying out compliance duties. This requires a shift from agency-based analysis to architecture-based analysis.

2. Recognising overlapping and shifting roles

In non-linear architectures, actors frequently occupy multiple roles simultaneously or transition between roles depending on the context. For instance, a user may act as a liquidity provider, a governance participant, and an end-user of the protocol. Validators may contribute to settlement without controlling custody. Developers may have contributed to the creation of a protocol but may no longer influence its operations.

The qualification of actors must therefore be dynamic rather than static. A one-time assessment of an actor’s role is insufficient; the relevant function must be analysed in relation to the specific interaction and the temporal context.

3. Differentiating between factual control and functional contribution

A crucial distinction must be drawn between factual control and functional contribution. For AML purposes, responsibility cannot be attributed solely based on the performance of a function; it must be based on the capacity to influence or direct the use of that function. Validators contribute functionally to settlement but do not control asset flows. Front-end operators facilitate interaction but do not control the protocol. Developers provide initial code but may not govern operational behaviour. A functional analysis must therefore assess both the nature of the function and the ability of the actor to shape its execution.

V. Challenges of AML Qualification in a Fragmented Ecosystem

The non-linear architecture of decentralised systems creates several specific challenges for AML qualification.

First, the absence of chokepoints eliminates the natural interfaces through which AML regulation traditionally applies. Without custodial institutions or centralised exchanges, regulators struggle to identify actors capable of performing monitoring, due diligence, or reporting.

Second, the fragmentation of functions prevents clear attribution. A DeFi transaction may involve smart contracts written by one group, governance parameters set by another, liquidity supplied by a third, and settlement performed by validators. Identifying a single actor responsible for AML compliance is conceptually difficult.

Third, the variability of user roles challenges the distinction between professional and non-professional actors. AML frameworks assume that regulated actors operate “as a business.” But many activities in decentralised systems—such as running a validator node or providing liquidity—may be performed by users without any business intent, sometimes as a technical necessity for the system’s operation.

Fourth, the cross-border nature of decentralised systems undermines jurisdictional coherence. AML regulation is territorial; blockchain networks are not. A functional analysis must therefore account for the extraterritorial and permissionless character of the ecosystem.

VI. Toward a New Paradigm for Actor Qualification in AML Law

Given the structural mismatch between AML assumptions and decentralised architectures, a new conceptual framework is needed. Rather than forcing decentralised actors into ill-fitting categories, regulators should consider an approach based on functional capacity and risk exposure, not institutional form.

This alternative model would evaluate actors according to their practical ability to influence value flows, their control over access interfaces, and their contribution to protocol-level functionality. It would recognise that some roles—such as front-end operation or token issuance—may be appropriate targets for AML obligations, whereas others—such as liquidity provision or participation in consensus—do not entail sufficient control to justify regulatory attribution.

Moreover, regulators should consider a shift towards protocol-embedded compliance, relying on cryptographic identity attestations, zero-knowledge proofs, or risk-scoring mechanisms that do not require centralised operators. This would align AML objectives with the technological reality of decentralised ecosystems.

Conclusion

The qualification of actors within crypto-asset systems cannot be achieved through the simple transposition of traditional AML categories. Decentralised architectures challenge the assumption that functions correspond to identifiable institutions and undermine the linear logic upon which AML regulatory design is based. A functional analysis must therefore be adapted to account for technological decentralisation, overlapping roles, and the dynamic nature of crypto-asset ecosystems.

The FATF’s functional VASP model represents an important doctrinal evolution, but it remains constrained by institutional analogies and an underlying commitment to centralised intermediation. A more nuanced and technologically grounded approach is required—one that acknowledges the autonomy of protocols, distinguishes between functional contribution and factual control, and develops compliance mechanisms compatible with decentralised structures.

Only by adopting such a paradigm can AML frameworks maintain their effectiveness without distorting the fundamental characteristics of the decentralised systems they seek to regulate.

Key takeaway. In non-linear, protocol-driven architectures, AML qualification must follow functions and control capabilities, not legacy institutional forms. A next-generation framework should pair functional analysis with protocol-native compliance tools rather than stretching the VASP concept beyond its doctrinal limits.