Can the CRA Collect for the IRS — or the IRS for the CRA? Article XXVI A, the Citizenship Carve-Out, and What Cross-Border Debtors Should Actually Expect

Most tax treaties stop at information exchange; the Canada-US convention goes further, and Article XXVI A is the reason cross-border debt advice keeps circling back to two questions — is the debt finally determined, and what was your citizenship when it arose. The machinery: a country may ask the other to accept its revenue claim for collection; an accepted claim is collected by the requested country as though it were its own tax debt — the CRA applying its collection powers (requirements to pay, garnishment, set-off, liens) to an IRS claim against Canadian-situs assets, or the IRS applying levies to US assets for a CRA claim — with the revenue remitted to the requesting country. The conditions that gate it: the claim must be finally determined — assessments with appeal rights exhausted or lapsed; a debt still in objection, appeal, or Tax Court is not eligible, which makes the domestic dispute calendar a collection-defense calendar too; the request must satisfy the article's procedural requirements and the requested country's acceptance; and the article covers taxes, interest, penalties, and costs across the treaty's scope, with the requested country's own limitation rules governing its collection mechanics. The carve-out that reorders everything: a country is not required to collect a claim against an individual to the extent the claim relates to a taxable period in which that individual was a citizen of the requested country (with parallel protection for entities on their status at the relevant time) — meaning the CRA generally will not collect IRS debts from someone who was a Canadian citizen when the liability arose, and the IRS mirror-protects those who were US citizens when their CRA debt arose. The precision matters and rewards timeline work: citizenship when the claim arose — the dual citizen from birth carries the shield in both directions for personal income tax debts; the American in Canada who naturalized after the debt years does not (Canadian citizenship acquired later doesn't retroactively protect earlier-arising claims); and establishing when a claim arose against when citizenship was held is genuine analysis in files with long assessment histories. What this means in operating terms, stated without folklore in either direction: for the protected (Canadian citizens with US debts from their citizen years, and the mirror), the treaty channel is closed — but everything else remains: US-situs assets meet ordinary IRS collection, refunds offset, passports certify for the US-citizen side, and the debt follows any future US ties; for the unprotected (the green-card-era debt of the later-naturalized, the US citizen in Canada without Canadian citizenship, entities outside the carve-out), Canadian or US assets are genuinely reachable through the treaty once claims finalize — the channel is used, particularly for substantial finalized debts, and the assumption that the border is a wall is simply wrong for this population; and for everyone, the article's finality condition makes engagement timing strategic — resolution channels entered before finality (agreements, offers, objections) shape whether a claim ever becomes an eligible request. The adjacent machinery completes the picture: information exchange (Article XXVII and FATCA's flows) means each authority can see cross-border assets regardless of whether it can reach them; and neither country's courts will directly enforce the other's tax judgments outside the treaty channel (the common-law revenue rule), which is precisely why the article exists and defines its boundaries.

Key takeaways

• The channel is real: finally-determined claims, accepted for collection, enforced with the requested country's full domestic toolkit — treaty collection is operating law, not theory, and substantial finalized debts are its caseload.
• Finality gates it: debts in live objection, appeal, or litigation aren't eligible — the domestic dispute posture is simultaneously the treaty-collection defense posture, and letting appeal rights lapse is what opens the door.
• The carve-out turns on citizenship when the claim arose: protection for the requested country's own citizens as of the debt's origin — birthright duals shielded both ways; later naturalization protecting nothing earlier; the timeline analysis (claim arose when, citizen since when) worth doing precisely in any real file.
• Protected doesn't mean safe: the shielded debtor still faces the creditor country's own reach — its-situs assets, offsets, passport certification, and any future footprint there; the treaty closes one channel, not the ledger.
• Unprotected means genuinely exposed: the non-citizen-of-the-requested-country's local assets are reachable once claims finalize — the planning constant for Americans in Canada without Canadian citizenship and Canadians in the US without American, in mirrored directions.
• Visibility is universal even where collection isn't: information exchange and FATCA mean both authorities see the cross-border picture — asset location is not asset concealment, and files should be built on that assumption.

Applying it to a real file

The analysis runs in four steps: characterize the debt (which country, which years, finally determined or still disputable — and if disputable, the dispute calendar is priority one); run the citizenship timeline (status in the requested country as of when each year's claim arose — naturalization dates against assessment years, with mixed multi-year debts sometimes splitting between protected and unprotected slices); inventory the exposure that exists regardless (creditor-country assets, offsets, certifications); and choose the resolution posture accordingly — the protected debtor negotiates with ordinary collection pressure in view, the unprotected debtor negotiates knowing local assets are ultimately in the conversation, and both do better inside arrangements than outside them, because the treaty channel, like every collection tool, is applied to the disengaged far more than to the enrolled.

Worked example

Three debtors, one article. Debtor one: a born-dual citizen in Oakville with a US$95,000 IRS debt from catch-up years — Canadian citizen when every claim arose: the CRA cannot be enlisted; her Canadian home and accounts sit behind the carve-out. Her file still resolves by installment agreement, because her US brokerage account, her future refunds, and her passport renewals were never behind anything — the shield covered one channel, and the agreement was cheaper than living around the others. Debtor two: a US citizen who moved to Vancouver in 2015 and naturalized Canadian in 2023, carrying IRS debts assessed for 2017-2020 — the claims arose while he was not a Canadian citizen: no carve-out; once finally determined, those claims are eligible for a CRA-collection request against his BC assets. The analysis lands during his (still-live) appeal window, which becomes the strategy: the dispute is pursued on its merits with the collateral knowledge that finality has consequences, and the resolved balance enters an agreement before any request could issue. Debtor three: a Canadian citizen in Florida owing the CRA C$210,000 from pre-departure years, appeal rights long lapsed — Canadian citizen when the claims arose, so the IRS mirror carve-out protects him from a treaty request against his US assets; his Canadian rental property and accounts, meanwhile, are the CRA's ordinary domestic targets, and the lien that arrives on the Toronto property the following spring illustrates the article's real lesson: the treaty decides which country's tools apply — it never decided that none would.

Official sources

"The Contracting States undertake to lend assistance to each other in the collection of taxes ... referred to in this Article as a 'revenue claim'," which "is finally determined when the applicant State has the right under its internal law to collect ... and all administrative and judicial rights of the taxpayer to restrain collection ... have lapsed or been exhausted." Under paragraph 8, "No assistance shall be provided ... to the extent that the taxpayer can demonstrate that ... where the taxpayer is an individual, the revenue claim relates to a taxable period in which the taxpayer was a citizen of the requested State." — Canada-United States Tax Convention, Article XXVI A, https://www.canada.ca/en/department-finance/programs/tax-policy/tax-treaties/country/united-states-america-convention-consolidated-1980-1983-1984-1995-1997.html

Taxpayers can set up "a payment arrangement ... you can afford"; on an unpaid debt the CRA can garnish amounts owed to the taxpayer and "use federal or provincial credits or benefits to reduce your debt, including: Goods and services tax/harmonized sales tax (GST/HST) credits [and] Future income tax refunds." — Canada Revenue Agency, Payment arrangements and debt collection, https://www.canada.ca/en/revenue-agency/services/payments-cra.html

Practitioner note

Article XXVI A is the collection folklore corrector in both directions: the border is not a wall for the unprotected, and the carve-out is not a pardon for the protected — it closes one channel while the creditor's own toolkit runs on. Every cross-border debt file we take starts with the same two timelines, finality and citizenship-when-arisen, because those two facts sort clients into strategies more decisively than the balance does.

See also: Browse every cross-border tax topic guide, organized by situation · Short version: Tax Debt Across the Border: Can the CRA Collect an IRS Debt, and Can the IRS Collect a CRA Debt?.

Next step

Fairlight prepares the collection exposure analysis — finality and citizenship-timeline determinations, the carve-out's application to each year's claim, the regardless-exposure inventory, and the resolution posture built on which tools actually apply. See cross-border pricing or book a call.