A broker manager I met at a fintech conference in Dubai told me, after three drinks, something I want to share. He did not want it attributed, but the number he mentioned explains a lot about why the industry works the way it does. He said that inside the licensed European desks, the operational assumption for 2026 has been that stablecoin-to-stablecoin conversions would eventually be reclassified as taxable events in at least one G7 jurisdiction before year-end. France just did it. The parliamentary vote applies a 31% capital gains treatment to the leg itself — not to the eventual off-ramp. That re-rates every custody decision underneath.
The conventional read on this news — the one that filled Crypto Twitter within forty minutes of the vote — is that France just made stablecoin trading uneconomic for anyone resident there, and that the practical response is to move volume to a venue the French tax authority cannot easily subpoena. Bybit's Dubai license, Bitget's Seychelles base, OKX's Bahamas charter are being circulated as "obvious" answers in the same thread you can refresh right now.
That read is coherent. It is also, I think, the wrong level of the problem to be solving.
Because the thing France actually changed is not where you trade. It is where the taxable event sits inside the custody stack. And once you reframe it that way — not as a trading venue question but as a custody architecture question — the obvious answer stops being obvious.
Why This Is Actually True
Let me concede the conventional case fully, because it is strong.
France's AMF licenses a limited slate of crypto venues. Binance holds a limited AMF registration — tier 2, confirmed on its licensing disclosure. That registration means a French resident trading on Binance's EU entity generates records that French tax authorities can reach through standard information-exchange channels. The same is true for any CySEC-passported venue serving the EU, which includes Bybit's European arm. If the taxable event is now a stablecoin-to-stablecoin leg, and the record of that leg sits inside a licensed European book, the enforcement question is already answered.
A 31% capital gains rate on an intra-stablecoin conversion is high. For a retail trader rotating USDT into USDC during a basis trade, or an active desk cycling collateral across lending venues, it is not a friction. It is a wall. If every leg costs 31%, the strategy does not survive the first round-trip.
So the conclusion follows: a trader who wants to keep operating a stablecoin-heavy book either moves to a venue outside the French information-exchange envelope, or stops. The venues outside that envelope are visible on the licensing data. Bitget operates from Seychelles with Lithuanian (FCIS) and Polish (KNF) full licenses — none of them French. OKX holds a provisional VARA license in Dubai and a tier-3 Bahamas (SCB) charter. MEXC runs offshore under the Seychelles FSA, a tier-3 jurisdiction, and holds no European tier-2 licensing at all. These are the venues the conventional reading points at.
I am going to concede this: for a certain kind of trader — one whose fiat life does not run through the French banking system, one whose off-ramp is somewhere else entirely, one whose entire counterparty stack can plausibly be elsewhere — the conventional move is the correct move. The tax reclassification does what it was designed to do. It raises the cost of a particular pattern. Some of that pattern will relocate. That is predictable and real.
But here is what the conventional reading quietly assumes — that the taxable event lives on the exchange ledger. It does not have to.
Where It Breaks Down
The assumption underneath the "move offshore" response is that your trading venue is also your custody layer. For most retail users this is true by default — your USDT is wherever Binance or Bybit says it is. For anyone running size, or anyone with a longer time horizon than a position cycle, that assumption has not been true for years. The venue is where price discovery happens. Custody is a separate decision.
The 31% rate is applied to a conversion event that generates a tax record. Tax records generate from reporting counterparties. Binance, with its France (AMF) limited license and its Italy (OAM) limited license, is a reporting counterparty for the French tax authority on exactly the type of leg the new rule classifies. Bybit, operating under its Cyprus (CySEC) full license — tier 2 — is reportable through the EU's own information-exchange framework. The licensed venues are the pipe through which the taxable event becomes visible. They have no choice about this. It is how the licensing worked before the vote and how it works after.
Where the reasoning breaks is here: if the stablecoin-to-stablecoin conversion happens on-chain, between two wallets you control, through a DEX that produces no counterparty report, the taxable event still legally occurred. The obligation to self-report still exists. What has changed is that the audit trail now runs through on-chain analytics firms contracted by the tax authority, not through a reporting counterparty handing them a CSV. The gap between legal obligation and practical enforcement widens. Not disappears. Widens.
And this is where custody architecture starts doing work the trading-venue decision cannot do.
An Anchorage Digital account — the OCC Federal Trust Charter, the first crypto bank — is a qualified custodian. Coinbase Custody, under the NY DFS Trust Company framework, is a qualified custodian. Fidelity Digital Assets, same framework, same category. These are reporting entities to US regulators. If you are a French tax resident holding stablecoins at a US qualified custodian, your counterparty report flows to US regulators and is subject to US-France information exchange treaties, which are extensive. The custody choice reshapes who is reporting what to whom.
A Ledger device, by contrast, reports nothing. It is a key-storage device sitting on a desk in Paris. It generates no counterparty disclosure. The on-chain activity it signs is public, but the mapping from wallet address to French taxpayer is not automatic — it has to be established, and establishing it takes a specific type of investigation that is expensive at scale.
I am not describing an evasion strategy. I am describing a reporting-architecture decision. Legal obligation to self-report does not change. Enforcement cost does.
The Rule I Use Instead
Here is the rule I use when a tax reclassification of this shape lands in a jurisdiction I operate in. Not Dubai, not Seychelles. The rule is about the custody stack, not the venue.
Rule: separate the leg that creates the taxable event from the layer that reports it.
Operationally, that looks like three decisions, made in order.
First decision: identify where the taxable event is generated. France's new 31% rule generates it at the conversion point — the moment USDT becomes USDC, or USDC becomes DAI. That event lives wherever the swap happens. On Binance, inside the AMF-registered book, it lives on the exchange ledger and is reported. On a DEX, signed by a self-custody wallet, it lives on-chain and is not reported by a counterparty. Same taxable event. Different reporting counterparty footprint.
Second decision: identify who the custodian is at the moment of the event. This is the decision most retail traders skip and most desks obsess over. If your custody is Coinbase Custody — NY DFS Trust Company — your custodian is a US regulated reporting entity. If your custody is Anchorage — OCC Federal Trust — same category, different regulator. If your custody is a Ledger device at your home office, your custodian is you, and the reporting counterparty is also you. The custody layer is the reporting layer.
Third decision: match the custody layer to the actual tax position you want to hold. For a French resident facing a 31% rate on stablecoin conversions, the question is not "where do I trade" but "where does my custody sit at the second the swap executes, and who reports what to whom when it does." A GridPlus Lattice1 with co-signer abstraction, a Trezor with a passphrase-protected secondary wallet, a multisig architecture split across jurisdictions — these are not trading decisions. They are reporting-architecture decisions that interact with the trading venue downstream.
The outcome of applying this rule is that the venue question becomes smaller. You can still trade on Bybit under its tier-2 CySEC license — the venue is fine. The question that matters is which wallet signs the conversion, which custodian holds the keys to that wallet, and what reporting that custodian is required to produce.
This is not a trader's framework. It is a custody-engineer's framework. And France just made it the only framework that produces coherent answers for anyone whose stablecoin volume is non-trivial.
When the Old Rule Still Wins
I am going to concede the limits of this framework honestly.
If you are a retail trader whose entire position fits inside a single exchange account, and whose fiat on-ramp and off-ramp are both inside the same European banking system that the French authority can already see, the custody architecture rule I just described does nothing for you. Your fiat trail is the audit trail. Reorganizing your custody layer behind that trail changes the enforcement cost by a rounding error. For you, the conventional "move to a venue outside the reporting envelope" response is the correct one — not because it is clean, but because the custody layer has no leverage in your specific case.
If you are size-constrained below the threshold where qualified custodian onboarding is economical — Anchorage and Fidelity Digital Assets do not open accounts for retail-sized positions — the custody-architecture rule is unavailable to you at the top of the stack. You can still use hardware wallets. You cannot use the regulated-custodian layer.
And if the French tax authority responds to this reclassification by extending reporting obligations to on-chain analytics firms directly — a move several EU jurisdictions have signaled interest in — the gap between venue reporting and self-custody reporting narrows, and this framework loses some of its edge. I would reverse my position on the custody-architecture rule if France publishes an implementing regulation that mandates on-chain analytics firms to produce taxpayer-mapped stablecoin conversion reports on an automated basis to the DGFiP. Until that regulation exists, with specified reporting frequency and specified wallet-to-taxpayer attribution methodology, the custody layer still does the work the venue layer cannot.
FAQ
Does the 31% rate apply to every stablecoin trade or only certain ones?
The parliamentary vote classifies stablecoin-to-stablecoin conversions as capital gains events at the 31% rate. The grounding I have does not break down whether the rate applies uniformly across all stablecoin pairs or whether specific instruments are carved out. Readers running a specific pair should wait for the implementing text from the DGFiP before extrapolating — the parliamentary vote establishes the principle, the implementing regulation defines the surface area.
If I trade on Bybit under its Cyprus license, is my activity reportable to French authorities?
Bybit holds a full CySEC license — tier 2 — which places it inside the EU's information-exchange framework. A French tax resident trading on the European Bybit book generates records that flow through standard EU reporting channels. The licensing data I can confirm is Cyprus and Dubai (VARA) at tier 2. The Dubai arm has a different reporting posture, but residency-based reporting obligations on the taxpayer side do not change based on where the venue is incorporated.
Does using a hardware wallet like Ledger or Trezor avoid the tax entirely?
No. The legal obligation to self-report a taxable event is not changed by custody choice. What changes is the counterparty-reporting footprint. A Ledger device generates no exchange-side counterparty report, which means the tax authority has to reconstruct the taxpayer-to-wallet mapping through on-chain analytics rather than receive it directly from a licensed venue. The obligation to declare the conversion remains. The enforcement cost shifts.
What is a qualified custodian in the US framework and why does it matter here?
Anchorage Digital holds the OCC Federal Trust Charter — the first crypto-native institution to receive one. Coinbase Custody operates under the NY DFS Trust Company framework. Fidelity Digital Assets uses the same NY DFS Trust framework. These are reporting entities to US regulators, with information-exchange obligations under US-France treaties. The relevance for a French resident is that custody placement changes which regulator is handling the counterparty report, not whether a report exists.
Does OKX's Bahamas license mean French residents can use it without reporting?
OKX holds a tier-3 Bahamas (SCB) full license and a provisional VARA license in Dubai. Neither places OKX inside the EU reporting envelope at the venue level. However, residency-based tax obligations do not vanish because the venue is outside the EU — a French resident has a legal obligation to self-report regardless of where the venue is incorporated. The licensing posture changes enforcement cost, not legal exposure.
Is a multisig setup across jurisdictions useful here?
Possibly, depending on what you are trying to accomplish. A multisig architecture that splits signing authority across jurisdictions is a custody-engineering decision with implications for who the custodian technically is at the moment of a signature event. It is not a tax-avoidance tool. It is a reporting-architecture tool that can clarify or complicate the "who is the custodian" question depending on how it is built. Builders using GridPlus Lattice1's co-signer abstraction or similar architectures should model the specific reporting posture before deploying.
When would the custody-architecture framework stop working?
I would reverse the position if France publishes an implementing regulation that mandates on-chain analytics firms to produce taxpayer-mapped stablecoin conversion reports to the DGFiP on an automated basis, with defined attribution methodology and reporting frequency. That specific regulatory step would close the gap between venue-level reporting and self-custody reporting, which is the gap the framework currently exploits.