Prof. Barry Appleton, Appleton’s Clause & Effect Substack Blog | Reading the Trade War | Part 3 of 3 | August 21, 2026
A concession you cannot deliver is a grievance with a delay on it.
The countdown that began on Tuesday night runs out at the end of the day today (Friday August 21). Whatever is signed, or is not, the question underneath it has had almost no public examination, and it is the one that decides whether any of these holds.
Can Ottawa deliver what Ottawa is promising?
On Wednesday afternoon the Prime Minister chaired a virtual cabinet meeting and then briefed the premiers. That sequence is worth pausing on. The terms had been agreed with Washington on Tuesday night. The provinces were briefed on Wednesday. A briefing is neither a negotiation nor consent. The order in which those two things happened is the whole subject of this piece.
Return to the second of the three questions any concession has to survive. Is it deliverable by the party actually promising it? Put that to the four headings Washington published on Tuesday night. The answer keeps coming back no.
The rule underneath this is old and settled. Ottawa may sign any treaty it likes. Implementing one is a different matter. Where the subject falls within provincial competence, Parliament cannot legislate it into force on its own. That has been the law since the Privy Council decided the Labour Conventions case in 1937, and no Canadian government has succeeded in dislodging it since. The federal government has real tools, including its spending power and its criminal law power, and it has used both to reach into provincial fields before. Neither of them puts a bottle on a shelf.
Before the individual files, look at the shape of the exchange as a whole.
What Canada stands to receive is relief on steel, aluminum, autos and forest products, and the reporting suggests it is partial. Steel and aluminum duties are said to fall from fifty per cent to twenty five. That is relief from a rate that did not exist eighteen months ago. Every one of those is an American tariff, imposed by proclamation and removable by the same instrument. Washington can deliver all of it unilaterally, this afternoon, with a signature.
What Canada is reported to be giving includes ending provincial restrictions on American alcohol and ending provincial procurement policies that exclude American suppliers. Neither is Ottawa’s to give.
Canada receives what the counterparty can deliver alone, and gives what Ottawa cannot deliver alone. That asymmetry is the structural risk in this agreement. It is visible before anyone has read a word of the text.
We now know how load-bearing those provincial decisions are, because a premier described the conversation, and the description is more revealing than anything either government has published about the substance of the agreement itself.
Manitoba’s premier told reporters what the Prime Minister had told the premiers. No alcohol back on provincial shelves, no trade deal. That was the message. Asked how hard the request was pressed, he said he would not call it begging, and then asked what the step before begging is called.
Read that back slowly. Ottawa has told the provinces that an international agreement it already announced depends on eleven separate decisions it cannot make. That comes from the Prime Minister’s own account, relayed by a first minister. No commentator had to read it into the constitution.
La Presse put it more bluntly than anything in the English coverage. The fate of a possible agreement rests in the hands of three provinces. Those three, Quebec, Ontario and British Columbia, are the ones the tariff war has hurt most, and they are the three that have declined to move.
Start with alcohol. It is the cleanest case. Eight of ten provinces have blocked American alcoholic beverages since early last year, and Washington named that conduct as a rationale for these tariffs. Liquor distribution in Canada is provincial. The boards are provincial Crown corporations operating under provincial statute, and the federal government cannot direct them to restock a shelf. Ottawa may promise comprehensive market access. Ottawa cannot put one bottle on one shelf in a province that has decided otherwise.
The digital heading is more complicated, and the complication matters, so it is worth being precise rather than rhetorical.
Privacy is shared, and the split runs differently from alcohol. It is worth noting first how little room the existing text leaves. CUSMA Article 19.12 forbids requiring that data be stored on Canadian servers. It runs to a single sentence and contains no public policy exception at all, unlike the equivalent provision in the CPTPP.
Within that room, the federal reach is substantial. The federal statute covers federally regulated businesses in every province, all commercial activity in the seven provinces without a substantially similar law of their own, and every interprovincial and international transfer of personal information regardless of where it originates. Only Alberta, British Columbia and Quebec maintain comprehensive private-sector regimes, and even there the federal statute governs cross-border movement. So if digital trade alignment is about data crossing the border, Ottawa holds most of that pen and can largely commit.
Artificial intelligence runs the other way, and this is the part almost nobody is examining.
Canada has no federal artificial intelligence statute. The proposed one died on prorogation in January 2025 and has not been reintroduced. What governs AI deployment in this country today is an assembled patchwork: the federal privacy statute and its provincial equivalents for the personal data inside the systems, Quebec’s Law 25 for automated decisions affecting individuals, a financial regulator’s model-risk guideline for federally regulated institutions, a Treasury Board directive covering federal government systems only, and human rights law for discriminatory outcomes however they are produced.
Read that list again and notice where the binding rules actually sit. The most demanding automated-decision requirements in Canada are Quebec’s, not Ottawa’s. Ontario legislated the disclosure of AI use in job postings effective this January through its provincial employment standards statute. British Columbia, Alberta and Saskatchewan have all moved on public-sector AI governance. The provinces are ahead of the federal government on the substance, and Ottawa has no statute to align.
So the question a negotiator should have to answer out loud is this: align what, exactly, and bind whom? Ottawa can commit to its own procurement rules and directives. It cannot commit Quebec’s automated-decision regime, or Ontario’s employment standards, or a federal AI statute that does not exist.
Procurement is the third fracture, and it cuts in Canada’s favour if anyone thinks to use it. Article 19.2(3) excludes government procurement from the digital trade chapter altogether, as a categorical scope exclusion carrying no necessity test. Provinces and municipalities buy on their own terms, largely outside the agreement’s coverage (which relies on a WTO agreement), and a federal promise of access for American suppliers applies only to what Ottawa itself buys.
That exclusion is the one door still open, which is why the thing to watch in Friday’s text is any language that qualifies it.
The door is already being examined. The American trade barriers report published in March named Canada’s sovereign cloud computing initiative among its complaints, alongside the Online Streaming Act and the Online News Act. Sovereign cloud is a procurement policy. It is the government deciding what the government will buy. Article 19.2(3) leaves that alone. Washington has flagged it anyway. It is also no longer a detail reported second-hand. Ottawa asking the provinces to end procurement policies that exclude American suppliers is now the headline.
There is a second version of the same difficulty, and it appeared this week from a premier who has agreed to comply. Manitoba’s premier said he was reluctantly open to restocking. In the same breath, he urged Manitobans not to buy the product. That is a lawful and honest position. It also shows the problem exactly. A government can deliver the form of a commitment while the result never arrives. The counterparty judges the result. Nobody grades the effort.
Premier Houston added a gloss worth pausing on. Provinces could still favour Canadian suppliers, he suggested, so long as they do not formally say so. That may be practical politics. As a matter of trade law, it is the least safe position available, because a preference that operates without appearing in the text is exactly what a disguised restriction on trade means, and it is far harder to defend than a preference stated openly and applied on neutral criteria.
Supply management is the fourth, and it may already show the argument working. Quebec has treated it as a red line throughout, and the federal trade minister has now said on the record that the supply-managed system will remain intact. A jurisdiction stated a constraint publicly and early. The file appears to have moved around it. That is what a stated constraint does when it is stated in time.
And this week the point stopped being theoretical. The two asks put to the premiers were specific: restock American alcohol, and remove anything from provincial procurement policy that specifically excludes American suppliers. Alberta, Saskatchewan and Nova Scotia responded warmly. Quebec, Ontario and British Columbia did not.
There is a further wrinkle that should trouble anyone drafting this text. The vice-chair of the premiers’ council has said that on Wednesday’s call every premier agreed to return the American alcohol. Several have not confirmed it publicly since, which is a gap worth noticing, because the difference between what is said on a call among first ministers and what a government will defend in public is exactly the space where implementation problems live. Agreement on a conference call binds nobody and creates no instrument. It is not what a trade partner relies on when somebody checks the shelves in six months.
Premier Christine Frechette put it about as plainly as a first minister can. Quebec, and Quebec alone, decides that question. Not Ottawa. She said she required further information from Ottawa before she could take a position, and that she had not received answers to her questions. On Thursday, she said the Prime Minister had answered many of them during a lengthy call, but she still stopped short of endorsing the agreement. Ontario’s position has been that American alcohol does not return without assurance the province’s manufacturing sector is protected. The premier has said nothing publicly since the briefing. British Columbia issued a statement expressing confidence that Ottawa had made substantial progress across strategic sectors, and did not say whether it would restock.
There is a parliamentary dimension as well. The leader of the Bloc Quebecois has formally written to the Prime Minister, asking to meet party leaders without delay to discuss the guarantees being obtained to protect Quebec’s interests. His stated concern is that Quebec should not become the currency of an agreement concluded in Washington. Whatever one makes of the politics, the request identifies the same gap this piece describes: the people who would have to live with the commitment have not seen it.
One thing this argument does not claim is worth stating plainly. Nothing here predicts that the provinces will refuse. They may well comply, particularly if the tariff relief is real and the pressure is sustained. Alberta and Saskatchewan never removed American alcohol at all.
The claim is narrower and survives either outcome. A commitment whose performance depends on eleven separate decisions is a different kind of commitment from one the signing government can perform itself, and it should be described that way in the text. If the provinces comply, nothing is lost by having said so. If they do not, everything turns on it.
So consider what a Friday signature produces. Ottawa commits. The provinces hold the constitutional authority that matters, and decline. Washington observes that the discrimination it named has not stopped. All three of those things can be true at once.
That last sentence is the dangerous one, and it is why this belongs below the waterline rather than in the political weather above it. Section 338 permits escalation where the discrimination continues, and the ceiling runs well past fifty per cent, all the way to exclusion of the country’s goods altogether. A federal promise the provinces will not keep does not merely fail. It creates the documented record of continued discrimination on which the next, larger instrument would be built.
We know this because Canada has already run the experiment once.
In the same agreement, Canada guaranteed its partners civil and criminal protection for trade secrets. The civil law of trade secrets is property and civil rights in the province, section 92(13). It is not Ottawa’s to write. Faced with a promise it could not fully keep, Parliament delivered the fraction it controlled. One Criminal Code offence, on safe federal ground, in the implementing legislation of 2020. The civil half was left as judge-made common law, where it sits today.
Six years on, the United States still lists Canada on its Special 301 Watch List for inadequate intellectual property protection. The treaty clause did not produce the protection. It could not, because performance required domestic law that Canada had not built and did not have the unilateral power to build.
Canada gave the concession and got the listing anyway. That is losing twice.
Someone will say a watch list is only a report and carries no sanction. That was the better view once. It is no longer safe. The same annual reporting process now feeds directly into action: the measures named in this year’s trade barriers report are the measures showing up in negotiations and, in one case, in tariffs. A listing carries no penalty of its own. It is the file the penalty gets built from.
The power hiding inside the limitation
Here is where the analysis usually stops. It should not.
Everything above reads as weakness. The instinct is to treat the federation as an embarrassment, a structural defect to be worked around, papered over, or apologized for in a negotiating room. That instinct is wrong, and it is expensively wrong.
Nobel Economics Prize winner Thomas Schelling made the point sixty years ago, and nobody has improved on it. The power of a negotiator often rests on a manifest inability to concede. A commitment is credible when the person making it cannot walk it back. The most credible constraint of all is one the other side can verify for itself.
Canada’s constraint is verifiable. It sits in the Constitution Act, 1867, a century-and-a-half-old document, predating every person now at the table. Any American trade lawyer can read it in an afternoon. That is a credible commitment device, and Canada owns one outright.
Washington uses its own version constantly and without embarrassment. Congress will not ratify that. Congress requires this report. Congress mandates that advisory committees be consulted before a joint review. American negotiators do not present congressional constraint as a failing. They present it as a boundary, and the boundary works for them.
Canada holds the same instrument and hides it.
Consider the two ways to handle the alcohol file. In the first, Ottawa promises comprehensive market access. The provinces decline. The promise fails, and that failure becomes evidence of continued discrimination under a statute whose next rung is exclusion. In the second, Ottawa says plainly and early that provincial liquor distribution is not within federal competence, that no Canadian government can commit it, and that here is what Ottawa can commit instead. The second produces a smaller agreement. It also produces one that survives contact with the constitution, and it keeps the constraint intact for the next round instead of spending it on a promise nobody could perform.
A reasonable objection at this point is that every federation faces this, and most manage. That is true, and the way they manage it is instructive.
The United States has the same difficulty. Its states regulate insurance, alcohol distribution, and much of procurement, and Washington signs trade agreements anyway. It does so by drafting for the problem. Sub-federal measures are addressed expressly. Obligations are scheduled against named entities. Reservations are listed rather than assumed. The technique is unglamorous and entirely available. Canada has simply not used it here, and the reason appears to be that using it would require saying out loud, in the text, that Ottawa’s writ does not run everywhere.
Treat the federation as the reason the work has to start earlier and be coordinated across eleven governments rather than announced by one. Alibi is the wrong word for it. That is slower, which is precisely why it must begin before a deadline rather than in response to a watch-list citation years afterward.
Build the capacity first. Then sign the clause. That sequence is the difference between a commitment and a wish.
A concession you cannot deliver is not a concession. It is a grievance with a delay on it, and the delay is running.
What Friday actually requires
Nothing in this argument counsels refusing a deal. It counsels signing one Canada can perform.
That means naming, in the text, which commitments bind Ottawa alone and which require provincial concurrence, and saying so to the counterparty rather than discovering it eighteen months later in a compliance dispute. It means bringing the provinces into the room before the signature rather than after, which is slower and is the only sequence that produces an agreement capable of surviving its own implementation. And it means treating the constitutional constraint as something to state plainly at the table, because a constraint the other side can verify is worth more than a promise the other side will eventually test.
Code before clause. Sign what you can deliver. Say out loud what you cannot.
The series
Reading the Trade War, three parts.
Four Trade Headings and a Countdown. What Washington published on Tuesday night, and what Ottawa did not.
The Trade Horizon Problem. What teaching sailing taught me about reading a trade war.
The Promises Ottawa Cannot Keep. This piece.
Further reading
The Clause Canada Could Not Keep, on what happens when Ottawa signs a treaty obligation the Constitution will not let it deliver.
A Sovereign Advisory System for Canada, Centre for International Governance Innovation, October 2025.
Whose Law Governs Canadian Data?, SSRN working paper, and the Balsillie Papers Special Report drawn from it.
The Cloud Casts a Long Shadow and The Digital Hinge of Sovereignty, on cloud law and the limits of data residency.
The Rules Washington Is Writing, on Canadian digital sovereignty under the CUSMA review.
The argument is developed at length in my forthcoming book, Own the Rails: Canada’s Fight to Decide Its Own Digital Future.
Prof. Barry Appleton is Interim Director of the Balsillie Legal Advisory Centre at the Balsillie School of International Affairs, Adjunct Professor at Wilfrid Laurier University; Managing Partner of Appleton & Associates International Lawyers; and Distinguished Adjunct Professor of Law and Co-Director of the Center for International Law at the New York Law School.
© 2026 Barry Appleton. All rights reserved.