A bullion purchase funded from digital assets runs through two record systems built to answer different questions. The digital-asset side generates addresses, transaction hashes, platform account statements and screening outputs — records that trace movement precisely and say little about origin. The bullion side generates contracts, invoices, allocation records and vault confirmations — records that establish ownership of identified bars and say nothing about where the money began. The institution receiving the fiat settlement sits between the two, and neither set answers the question it is obliged to ask.
The failure is consistent in shape. The counterparty produces a signed message proving control of an address, a screening result, an account statement, sometimes all three, while what stays open is provenance: how the assets were acquired and how they moved before they became fiat. Proof of control establishes who can operate an address at a moment in time; it does not establish how the assets in it were obtained. A reviewer who asked the second question records the first as a gap, and the settlement stays held while the request is reissued.
Since 1 January 2026 the same disposal is recorded a second time on a separate track — reporting crypto-asset service providers in first-wave jurisdictions now collect transaction and tax-residence data for automatic exchange between authorities from 2027 — so a file assembled today has to reconcile with a report the counterparty never sees. This page sets out what a gold purchase funded from digital-asset proceeds leaves in evidence: what a licensed digital-asset platform screens before it settles, what a named order reference binds together, and what the allocation documents establish about title to specific bars.
Why a receiving bank treats crypto-sourced funds differently
The scrutiny is disproportionate to the base rate, and the reason is in how the numbers are built. Chainalysis put illicit inflows for 2025 at a minimum of $154 billion, a 162% rise on the prior year, while its estimate of the illicit share of all attributed on-chain volume stayed under 1%. A compliance function does not work to the ratio. It works to the absolute figure and to the concentration behind it, because a single mismatched counterparty in a file is a supervisory finding regardless of what share of global volume it represents. TRM Labs measured 2025 differently, reporting that illicit entities captured 2.7% of available crypto liquidity — a metric framed against deployable capital. Two credible measurements of one year diverge by construction, and the reviewer holds both.
Asset choice moves the review before anyone reads the file. Stablecoins now carry 84% of illicit transaction volume, having become the default settlement unit for the same reasons legitimate users adopted them: low volatility and cross-border transferability. The operational consequence runs against intuition — an order funded in USDT or USDC draws more questions than a bitcoin-funded order of equivalent size, because the concentration sits on the rail the counterparty chose for its convenience.
What the institution is assessing is narrower than the asset class. It is assessing whether the origin of value can be evidenced to a standard its own supervisor will accept on inspection, and an incomplete record escalates on that ground alone. The only available response is to close the file — assembling one layer deeper the banking and settlement evidence the institution expects on a wire-funded purchase, because the value began on-chain. Transfer history counts for as much, and a leg that never passed through an obliged entity is categorised as elevated risk by policy, ahead of any fact about it. In the FATF’s 2026 survey, 88% of responding jurisdictions classified peer-to-peer transactions as high money-laundering, terrorist-financing or proliferation-financing risk.
Four questions, four artefacts: control, provenance, screening, identity
Four distinct questions attach to a crypto-funded settlement, and each has a different artefact. Offering one where another was asked for reads as a gap, not as partial compliance.
| The question | What it establishes | What answers it | What it leaves open |
|---|---|---|---|
| Who can operate this address? | Control at a point in time | A signature over a challenge message from the address’s private key, or a micro-transfer originated from it | How the assets in it were obtained |
| How were the assets obtained? | Provenance | Acquisition records — platform account statements, purchase confirmations, invoices for services settled in digital assets, disposal records of an earlier position | Whether the counterparty controls the address today |
| What risk is visible on-chain? | Exposure of the address and its counterparty clusters | A screening output produced by an obliged entity running analytics under its own authorisation | Provenance; a screening result is an assessment as at a date, not a history |
| Who is the person or entity? | Identity and beneficial ownership | The KYC or KYB file — incorporation documents, ownership register, authority to act | Any connection between that identity and these assets |
Of the four, control is the artefact most counterparties already hold, and the one most often mistaken for the whole answer: a valid signature is available for an address holding third-party transfers, stolen assets, company funds, borrowed assets or value received minutes earlier. The reason it is the artefact on hand is regulatory. The EBA’s Travel Rule Guidelines require proof of ownership where a transfer to or from a self-hosted address reaches EUR 1,000, alongside whitelisting controls — so a counterparty who has moved value through an authorised service provider in the EU has usually been asked to sign something, and arrives holding exactly one of the four artefacts.
Where the counterparty is a company, the substitution breaks further. A signature demonstrates that one natural person can access one address. Authority to sign and beneficial ownership are separate questions, answered by authority documents and the signing structure alongside the technical proof; where a multisignature threshold applies, the file has to show it was satisfied. Phrasing that asserts verified legal ownership or a clean address claims more than any method delivers, and invites the reviewer to test it; a file that names the person, the address, the method and the date survives that test. The signature answers for access, and provenance stays with the acquisition records.
The chain from acquisition to allocation, and where it breaks
A reviewer reads the chain as continuous or does not accept it. Six links run from the original acquisition to the allocation of identified bars, and the file is assessed at its weakest joint, whatever the volume of documents behind it. Most discontinuities occur at four of them.
Acquisition and holding records
Furthest from the settlement, the first link is also the hardest to reconstruct. A position purchased through a platform account produces a statement; one acquired as payment for services is evidenced by invoices; one built over years across several platforms survives as fragments. Where value has moved between self-custody addresses, across several platforms and over multiple chains, the history is transparent in technical terms and difficult to explain in compliance terms — institutions want a coherent chain connecting identity, ownership, acquisition and transfer, and transaction hashes supply only the last of those four.
Age works against the counterparty here. The platform holding the acquisition record is a commercial entity with a finite lifecycle, and a statement that could have been exported on demand three years ago may now sit behind a wound-down service or a closed account. Nothing on-chain replaces it.
Self-hosted addresses in the chain
Any period during which value sat outside an obliged entity is a period no third party attests to. That gap is bridged by three records the counterparty holds itself: the outbound transfer recorded by the originating platform, the inbound transfer recorded by the receiving platform, and the on-chain record connecting them. What the reviewer is asked to accept is that the address between the two ends was the counterparty’s throughout — and a signature demonstrates control on the day it was signed, leaving the interval to the transfer records alone.
For this leg the standard is under active revision. A file built to the current minimum may be read later against a tighter one, so the counterparty retains the full record of a self-hosted leg, beyond whatever a service provider asked for at the time. Article 75(2) of the EU Transfer of Funds Regulation required the Commission, after consulting the EBA, to report by 1 July 2026 on the risks posed by transfers to or from self-hosted addresses and to propose amendments where appropriate.
The conversion event
Conversion is where the two record systems touch, and the licensed digital-asset platform performs it against the permissions it holds. Golden Ark Reserve receives fiat, holds no digital assets and issues no token or claim on metal — so the conversion record originates with the platform and enters the file as the platform’s document, dated and scoped, not as a reconstruction by the seller.
Fiat settlement against the order reference
From quote to Evidence Set, one order reference attaches to the pro forma invoice for the advance payment, the gateway payment, the settlement and the allocation. That binding is what allows an incoming settlement to post against a specific order. Without it the payment arrives as an unattributed credit — a sum in a bank account with no document connecting it to a purchase, which is the same defect the counterparty came to solve, reproduced one step further down the chain.
What the licensed platform screens, and under whose licence
Two authorisations touch the purchase and they do not overlap. Screening of the address, the transaction and the on-chain counterparty exposure sits with the licensed digital-asset platform, under its own licence, using its own analytics, at the point of conversion. What reaches the seller’s file is that conclusion, with its date and its scope. That document travels as the platform issued it.
The perimeter is drawn around the handling of virtual assets, and the seller’s jurisdiction draws it explicitly. Oman requires virtual asset service providers to register with the regulator and to apply AML/CFT measures under Decision No. E/35/2023, with crypto activity otherwise captured by the country’s AML/CFT rules under Royal Decree No. 30 of 2016. The screening obligation and the licence to discharge it sit on the same side of that line.
For the counterparty this settles a question that otherwise stays open until the worst moment. The seller does not vouch for on-chain provenance to a receiving institution, because the seller never held the assets and never ran the analysis. It records what it received, against which order reference, and what it allocated. A counterparty planning to lean on the bullion documents as an answer to a provenance question is planning to answer with the wrong artefact.
What changed in 2026: parallel reporting and a closing record window
Two developments this year alter what a file has to survive.
Where a counterparty transacted through a reporting service provider, the disposal is now reported independently. The OECD framework went live on 1 January 2026 for the first-wave group — the entire European Union through DAC8, the United Kingdom, Canada, South Korea and Japan — with first reports to domestic tax authorities due in 2027 and cross-border exchanges following; Switzerland, Singapore, the United Arab Emirates, Hong Kong and Turkey target first exchanges in 2028, and the United States has committed to 2029. Reporting service providers verify tax residence, collect the taxpayer identification number and report annual activity — buys, sells, swaps and transfers. Retail payment transactions above a USD 50,000 threshold fall under dedicated reporting rules, with aggregation applied in defined cases. As of 23 June 2026, 46 jurisdictions had committed for the 2026 reporting period and 29 for 2027.
What this forces is consistency of record. A disposal described one way in the file handed to a seller and its bank, and reported another way by a service provider under an obligation the counterparty does not control, produces a discrepancy that surfaces years after the settlement cleared — when the underlying platform statements are hardest to retrieve.
Against that, retrieval itself got harder. The MiCA transitional period expired on 1 July 2026, with no further extension available: an entity providing crypto-asset services to EU clients without authorisation is in breach. Against an estimated pre-MiCA population, roughly 18–21% had become MiCA-authorised by late June 2026, and ESMA had told providers to have wind-down plans ready for implementation ahead of the deadline. Platform accounts closing under a wind-down take their statement history with them, and the acquisition record is the one link in the chain that cannot be rebuilt from on-chain data, which never carried an identity. Export precedes need.
Coverage of the transfer legs is improving but incomplete. The FATF reported in July 2026 that 83% of surveyed jurisdictions had passed Travel Rule legislation, up from 73% a year earlier, while 22% of assessed jurisdictions remained non-compliant and only one was fully compliant with Recommendation 15. A leg routed through a jurisdiction where the rule is legislated but not operating produces no counterparty-side record at all. The counterparty supplies that leg itself or it stays unevidenced.
What the purchase itself puts in the file
Alongside those, the transaction generates a distinct record set, and each document in it does one thing. In the order they arise:
- Master gold sale-and-purchase agreement — the contract between the counterparty and the Operating Entity, naming the parties and fixing the terms the rest of the file operates under.
- AML/KYC record — identity and beneficial ownership of the counterparty, sanctions screening against OFAC, EU, UK and UN measures, and the eligibility determination that follows.
- Pro forma invoice for advance payment — the sum requested against a named order reference, issued before specific bars, serial numbers and a final price exist.
- Payment confirmation — the fiat settlement received against that reference.
- Commercial invoice — issued against the bars actually allocated, and only after allocation.
- Allocation Record — the bars held for the counterparty, identified by serial number, weight, fineness, refiner and place of holding. Whole bars only.
- Confirmation of title to the bars — the counterparty’s ownership of those identified bars.
- Brink’s placement confirmation — placement in a dedicated client sub-account at Brink’s Hong Kong or Singapore, named to the counterparty in the vault register, segregated from Golden Ark Reserve’s own stock and from other participants.
- Instruction log — the sequence of instructions and hold events across the lifecycle.
One document the counterparty does not receive is the sanctions screening report. What enters the client-facing record is the Verified status, its date and an internal reference. The reason is the use a released report would be put to: handed over, it travels to a third party as a representation about a named person, made by an entity that ran it to decide on admission and warranted nothing to anyone else. The status and its date convey the fact without the exposure.
Nothing in this set speaks to acquisition. The nine documents begin at the contract and establish identity, settlement and title; the platform statements, purchase confirmations and transfer records that show where the value came from are generated years earlier by other parties and remain the counterparty’s to hold and produce. The AML/KYC controls applied at onboarding determine admission to the relationship, and the route carries a minimum order of 1 kg — which sets the scale at which this documentary weight is proportionate. What the counterparty holds at the end is allocated title to identified bars, recorded by serial.
How the package reads end to end — and the boundary
From the settlement, a reviewer works backwards. The fiat arrives with a reference; the reference resolves to an order; the order resolves to a pro forma, a conversion and a settlement performed by the licensed platform; the platform’s screening output carries a date and a scope; the counterparty’s own acquisition records trace the position back to its origin; the identity file names who was party to all of it. Where each link is present the file reads in one pass. Where one is absent the reviewer stops there, and the strength of the rest makes no difference to the outcome.
Because identity, provenance, screening and allocation have separate owners, no party can supply another’s. The seller documents what it received, against which order reference, and what it allocated. The licensed platform documents what it screened and what it converted. The counterparty documents where the value came from. A file assembled on the assumption that the bullion documents will answer a provenance question fails at the point it was assembled to survive.
The boundary is on record. Golden Ark Reserve makes no representation to a counterparty’s bank and issues no opinion on the tax treatment of a disposal in any jurisdiction; the Evidence Set is a set of records, and it is read by others on the terms it states. Counterparties working through the documentary requirements before committing to a purchase can review the funding route and its operating conditions in full.