On October 16, the French National Assembly advanced a provision that would make stablecoin swaps a taxable event, effective January 2027, conditional on the bill becoming law. That sentence contains six verifiable claims. I checked all six. Every source field in the underlying report was empty. There is no tax rate. There is no tax base. There is no amendment number. There is no vote tally. There is no link to the text.
I have spent nine years reading documents written to be believed rather than checked. Whitepapers, audit scopes, post-mortems, governance proposals. The pattern is stable: the artifact with the most missing fields is usually the one that matters most. Not because the authors are hiding something, but because the fields that would make a claim falsifiable are precisely the fields that require someone to have done the work.
So let me be explicit about what this piece is. It is not a report on a tax rate, because no rate exists in the source. It is an audit of a structural claim: that a legislature has decided to reclassify a transaction type, and that the reclassification is scheduled to coincide with the arrival of the data infrastructure that makes it enforceable. Everything I assert about mechanics is inference, and I will mark it. Everything I assert about the source's gaps is fact.
The code speaks louder than the whitepaper. Here, the amendment text speaks louder than the press release. We do not have the amendment text. That is the first finding.
Context: the seam France is aiming at
France has taxed crypto since 2019 under a flat levy on gains realized when crypto is converted to fiat currency. The architecture matters more than the number. Conversions between crypto assets have historically been treated as deferrals โ no taxable event, because in the legislator's model nothing had left the crypto perimeter. Value moved from one token to another; the tax claim followed the value and waited.
That deferral is a seam. In security terms, it is an unauthenticated boundary: a place where state changes occur without the system observing them. A taxpayer can rotate between assets indefinitely, realize nothing in fiat terms, and reset cost basis through the rotation itself. For a tax authority, this is not a loophole in the moral sense. It is a structural observation gap, and observation gaps are what tax authorities exist to close.
Now layer on MiCA. The EU's Markets in Crypto-Assets regulation creates legal categories, and the relevant one here is the e-money token. An EMT is defined by its claim: redeemable at par, on demand, against the reference fiat currency, issued by an authorized e-money institution or credit institution, backed by reserves with disclosure obligations. It is, functionally and legally, the closest thing in the crypto perimeter to money.
That is the coupling point. A token legally defined as a claim on fiat is the easiest token to argue is fiat for tax purposes. Once a legislature accepts that framing, the swap of an EMT for another asset stops looking like an internal transfer within the crypto perimeter and starts looking like a monetization โ a disposal into something that behaves like currency. The deferred treatment collapses.
The vehicle matters too. This arrived inside France's annual budget bill, the projet de loi de finances. The PLF is not a crypto bill. It is a financing instrument, and amendments ride on it the way payloads ride on a legitimate process. Its fate is collective: the provision lives or dies with the entire budget, and the budget runs the full procedural gauntlet โ Senate review, a joint committee if the chambers disagree, the government's constitutional power to force passage, and constitutional review before promulgation. A first-reading vote in the Assembly is the beginning of that sequence, not the end of it.
And then the date. January 2027.
Core: what the proposal actually does
The operation is taxonomy, not rate
Read the mechanism carefully. The proposal does not invent a new tax. It does not announce a rate. What it does is redraw the boundary of the taxable event: a stablecoin swap moves from the deferred bucket to the immediate bucket. That is a classification operation.
The distinction is not academic. A rate is a scalar โ you can model it, argue about it, price it. A boundary is a topology. Boundaries create disputes at the edges, because every boundary has a definition, and every definition has a case it did not anticipate. In nine years of contract review I have never seen a protocol fail because a constant was set wrong. I have seen them fail because a boundary condition was never stated. The integer overflow I found in the Zeek Token sale contract in 2017 was not a missing check in the obvious sense; it was a case the specification had simply not enumerated, and fifteen senior developers had read the same specification and assumed the enumeration was complete.
So the first question about this proposal is not how much. It is where the edge sits. The answer, from the source, is: the edge is EMT. Which immediately raises the cases the source does not address.
Is a bridged EMT the same asset as a native EMT? If USDC moves from Ethereum to Arbitrum through a native bridge, has a disposal occurred? If a wrapped representation is exchanged for the canonical representation of the same underlying claim, is that a swap of two assets or a change of custody for one asset? If an EMT enters a liquidity pool and later exits, is entry a disposal, exit a disposal, or both? If a user supplies EMT as collateral and is liquidated, does the liquidation constitute a taxable event, and at what price?
None of these are exotic. They are the ordinary vocabulary of anyone who has used DeFi for a week. If the taxable event is defined at the level of EMT exchange and no exemption is written for same-asset cross-chain movement, the proposal does not tax gains. It taxes motion. And a tax on motion is not a tax on income; it is a toll on architecture.
Why EMT is the right anchor for a tax authority
There is a reason the proposal anchors on MiCA's EMT category rather than on stablecoins generically, and it is not ideological. It is operational.
An EMT has an identifiable legal issuer. It has reserve disclosure obligations. It has an authorized institution behind it, subject to supervision. It has, in practice, a recognizable set of contract addresses. Compare the alternative: an arbitrary ERC-20 token with an anonymous deployer, no reserve attestation, and a contract that may or may not be a proxy pointing at code that changed last week. A tax authority cannot build a reporting regime on the second object. It can build one on the first.
This is enforceability by design, and it is the structural cost of MiCA compliance that nobody priced. Compliance makes you visible. Visibility makes you taxable. Every issuer that spent 2023 and 2024 lobbying for MiCA clarity was, in the same motion, volunteering to become the most legible target in the crypto perimeter. That is not a betrayal by regulators. It is the arithmetic of legibility. Trust is a vulnerability vector, and legal recognition is a specific instance of it: you cannot be trusted by a system without being enumerated by that system.
The corollary is an arbitrage gap the legislator almost certainly knows about and has chosen to accept, at least at this stage. If EMT swaps are taxable and non-EMT stablecoin swaps are not, the rational French taxpayer moves to the non-EMT instrument. If the friction is large enough, the flow goes to algorithmic designs, to non-compliant issuers, to offshore venues, to self-custody and DEX routing. The direction of regulatory arbitrage is usually opposite to the direction of legislative intent. A rule designed to capture stablecoin activity by making the compliant version taxable will, at the margin, select for the non-compliant version.
The missing variable that makes modeling impossible
Here I have to be blunt, because this is the part the coverage skipped.
The source gives me a taxable event and an effective date. It gives me no tax base and no rate. Those are not details. They are the two parameters that determine whether this is a technical patch or a substantive tax increase, and without them every downstream claim is unfalsifiable.
Consider the range. If the base is the gain on the swap โ disposal value minus adjusted cost basis โ then this is a scope extension of the existing capital gains framework, and the burden is administrative: you need cost basis continuity across assets, across chains, across time. If the base is the gross notional โ the full value of the EMT exchanged โ then at a flat levy in the range France applies to crypto disposals, a single round-trip swap could consume a third of the principal. At that point the provision is not a tax. It is a prohibition with a receipt.
These two readings produce opposite economies. One permits stablecoins to function as a medium of exchange with a compliance overhead. The other ends stablecoin usage in France for any purpose other than holding. The source does not distinguish between them. A reader cannot know which law is being proposed, and that is not a rhetorical flourish โ it is a defect in the information, and it should be treated as one.
A second-order problem follows. If the base is the gain, the regime requires a cost basis tracking standard for on-chain events at a granularity French law does not currently provide. Multiple swaps in a single block, gas paid in a third asset, rebasing instruments, LP tokens whose redemption value is not a price โ these need rules. Without them, sequential taxation of the same economic value becomes not a risk but a certainty. The same euro of value, rotated five times through EMT pairs, would be taxed five times on a base that was never adjusted. That is not a high tax rate. It is a broken accounting model, and broken accounting models are how systems lose integrity without anyone making a decision.
The observation gap: CEX sees, DEX does not
Now the execution layer, where I spend most of my professional attention.
A taxable event defined on EMT exchange is only collectible where exchange is observable. Centralized venues in France can be required to report. Custodians can be required to report. Authorized EMT issuers hold redemption records. But an on-chain swap through a DEX, a cross-chain bridge, or a non-EU venue with no French nexus is not observable by the tax authority except through reporting rails that are still being built. The result is a two-tier system: friction on the surface the state can see, flow to the surface it cannot.
This is a pattern I recognize from protocol design. You can only enforce invariants on state you can read. Every system that tries to govern an unobserved boundary ends up governing the observed side of it harder, which pushes activity across the boundary, which shrinks the observed base, which increases the pressure to govern the observed side harder still. Complexity is the enemy of security, and a rule enforceable only against the compliant half of a market is not a rule โ it is a selection mechanism for non-compliance.
January 2027 is the tell
The effective date is the most informative field in the entire source, and it is informative precisely because it looks administrative.
DAC8 โ the EU's eighth directive on administrative cooperation in tax matters โ requires member states to transpose by the end of 2025, applies from January 2026, and produces its first reporting cycle in 2027 covering 2026 data. The OECD's Crypto-Asset Reporting Framework targets first information exchanges in 2027 among participating jurisdictions. Both timelines converge on the year the French provision is scheduled to take effect.
That is not a coincidence, and the inference is clean: the legislator is waiting for the collection infrastructure before switching on the liability. First the data, then the enforcement. This also explains the roughly fifteen-month adaptation window. It is long enough for exchanges to build reporting interfaces, for EMT issuers to standardize redemption records, and for tax software vendors to ship a product.
Which brings me to the one genuinely positive commercial signal in this story, and I want to frame it carefully. The compliant tax-infrastructure layer โ tools that reconcile on-chain events into a declarable basis โ has a deterministic demand increase through 2027. That is a real transmission chain, unlike everything else in the source. But it is a second-order commercial opportunity, not a technical breakthrough. And I would flag a specific risk in it, drawn from an audit I ran last year.
In 2025 I reviewed an AI-assisted audit tool used by a mid-sized firm. The finding was not that the model was wrong. It was that the model was trained on historical compiler behavior and had no representation of vulnerabilities introduced by newer compiler versions. The tool was accurate about the past and blind to the present. Tax classification software built to 2026 specifications will have exactly this property with respect to 2027 instruments. Automation does not remove bias. It freezes bias into infrastructure and then charges rent on it. Every firm building the reporting layer for this regime should assume the definitions will change after the software ships, and should price the maintenance accordingly.
The procedural gauntlet nobody priced
The source reports a first-reading passage. In French budget procedure, a first-reading passage of a PLF is close to the default state โ the government generally has the votes, and the Assembly vote is the opening move. The provision then goes to the Senate, which has a long record of modifying or deleting tax increases. If the chambers disagree, a joint committee attempts a compromise; failure there can lead to a second reading, or to the government invoking the constitutional mechanism to force adoption without a vote. Then constitutional review, where the Conseil constitutionnel examines procedure and substance, including equality before the tax burden and the intelligibility of the provision. Then promulgation. Then, and only then, January 2027.
MPs back is a procedural statement, not a legal one. The probability that this specific clause survives in this specific form is not high, and no one reporting on it has the vote count to say otherwise.
The constitutional exposure deserves its own line, because it is the kind of thing that kills provisions quietly. If EMT swaps become immediately taxable while exchanges between other crypto assets remain deferred, the law treats two economically similar transactions differently. French constitutional review takes equality in taxation seriously, and the government's defense โ that EMTs are legally money-like and other tokens are not โ is coherent but not obviously sufficient, because the deferral for other assets is itself a choice, not a necessity. And if the base is gross notional rather than gain, the confiscatory character of the levy becomes a separate and stronger ground for challenge.
Contrarian: what the bulls get right, and what both sides miss
The reflexive industry response is that this is hostile. I think that reading is lazy, and I want to give the opposite case its strongest form, because the strongest form is where the real risk hides.
Taxation is the price of legal existence. No legislature bothers to define the taxable event for an asset class it intends to prohibit. The fact that the French state is arguing about how to tax an EMT swap rather than whether the instrument may exist is a concession of legitimacy. It is also an implicit admission that EMTs function as money โ precisely the argument the industry has spent years making and which MiCA's category structure now formalizes. The provision is, in a narrow sense, a compliment.
There is a stronger version. The real target may not be stablecoins at all. It may be the deferral seam itself โ the fact that rotation between crypto assets resets cost basis without a taxable event. The stablecoin swap is simply the most liquid, most abused, most legally legible route through that seam. If that is the intent, then the stablecoin framing is packaging, not payload, and the provision will expand in later readings to cover the whole rotation surface. That is the scenario worth tracking, and it is the one that would convert a localized friction into a national competitiveness event.
But both the bulls and the bears are arguing about a number that does not exist. The bulls say the friction is manageable. The bears say it drives activity offshore. Neither can be evaluated without a rate and a base, and neither has been published. Meanwhile the market has priced this at essentially nothing โ correctly, in my view, given the three-way attenuation of a first-reading vote, a 2027 effective date, and a scope limited to one token category. Volatility is just unaccounted-for variables, and here the variables are so thoroughly unaccounted for that the instrument cannot be priced at all.
Takeaway
The question is not whether France will eventually tax the stablecoin swap. The direction is set, and the 2027 synchronization with DAC8 and CARF tells you the state intends to see the transaction before it taxes it.
The question is whether anyone โ taxpayer, exchange, treasury โ will be able to compute the base.
Watch three fields. The amendment number, so the text can be read rather than reported. The base definition, gain or notional, because that single choice determines whether this is an administrative patch or a functional ban. And the cost-basis continuity rule, because without it the regime does not tax income, it taxes motion, and it will do so repeatedly on the same euro.
If those three fields are still absent when the Senate takes up the text, do not read the silence as an oversight. Read it as the finding. Every artifact is a trace of failure โ and a draft with no parameters is the trace of a legislature that has not yet decided what it is taxing.