Canada banned more than 1,500 assault-style firearm models in 2020 and promised compensation — the individual declaration window closed in 2026 with 68,717 firearms declared against a federal estimate of 144,000, and the amnesty now ends 90 days after a Supreme Court case that has not been heard

What is happening?

On 2020-04-18 and 2020-04-19 an attack in Nova Scotia killed 22 people, the deadliest such event recorded in Canada. Twelve days later, on 2020-05-01, the federal cabinet prohibited more than 1,500 models and variants of what it called assault-style firearms, among them the AR-15 and other commonly owned semi-automatic rifle patterns. It did this by Order in Council under the Criminal Code rather than by a vote in Parliament, and the prohibition took effect the same day.

The prohibition arrived with two promises attached. The first was an amnesty, so that people already holding the newly prohibited firearms would not be prosecuted while they waited, initially for two years. The second was a compensation program that would buy the firearms back. The prohibition took effect immediately. The compensation program did not.

Six years later the collection figures are far below the federal estimate of what is out there. A Public Safety Canada transition binder dated 2023-02-14 put the number of firearms in scope at about 144,000, counting only those previously registered as restricted, and recorded that the industry association put it at about 518,000 once unregistered firearms were included. The individual declaration window ran for ten weeks, from 2026-01-19 to 2026-03-31, and 37,869 people declared 68,717 firearms. That is less than half of the lower of the two estimates and about an eighth of the higher one.

The deadline that would have forced the issue has been moved four times, and the fourth time it stopped being a date. The amnesty was extended in March 2022 to 2023-10-30, in October 2023 to 2025-10-30, and on 2025-10-10 to 2026-10-30. Then an order registered on 2026-06-05 replaced the fixed expiry in all three of the earlier amnesty orders with a different rule altogether: the amnesty now runs until 90 days after Supreme Court of Canada files 41858, 41859, 41860 and 41861 conclude. Those files are a constitutional challenge to the prohibition itself, brought by a firearms-owner advocacy organization, dismissed at the Federal Court in October 2023 and at the Federal Court of Appeal in April 2025, and granted leave to appeal on 2026-03-19. The question in them is whether the cabinet could create a prohibition of this size by order rather than by statute. No decision has issued and reporting places one in the following year.

Meanwhile the prohibition itself kept growing. The list was extended again in late 2024, and by 2026-03-19 it covered more than 2,500 models, about 67 percent more than the order of 2020 carried.

Whose problem is this?

The body that made the prohibition is the same body that decides when the shield over non-compliance lapses, and it has moved that date every time it approached. Nothing outside that body sets the clock, and since June 2026 the clock is not a date at all.

Where does this problem end?

The boundary matters because the prohibition and the compensation program are two separate things and only one of them has actually been carried out. The prohibition is in force, it has been in force since the day it was made, and it has since grown by about two thirds. What has not happened is the part that was supposed to make the prohibition real on the ground, and the amnesty is the instrument that has kept the difference from surfacing as a criminal matter.

What is the state now, and what should it be?

Now

The state the program describes is one in which no prohibited firearm remains in private hands and the amnesty lapses because there is nothing left for it to shield. That end state has never been in dispute and it is what the compensation program exists to reach.

What is missing is a date. Until June 2026 there was one, and it had already been moved three times. The order registered on 2026-06-05 removed it and put a court process in its place, so the moment at which the prohibition becomes enforceable against ordinary holders is now set by four case files rather than by a calendar. No source opened here reports a federal target for how many firearms should have been collected by any particular date, nor a target rate of participation against either of the two estimates of the stock.

How big is it?

What is counted is firearms, and the counts do not agree with each other. Public Safety Canada put the stock at about 144,000 as of 2023-02-14, counting only firearms previously registered as restricted. The industry association put it at about 518,000, adding those never registered, and that larger figure appears in the federal costing of 2021 and in the federal binder of 2023 without either document adopting it or refuting it. The two differ by a factor of about 3.6, and no independent study opened here closes the gap.

Against those estimates, 68,717 firearms were declared by 37,869 individuals in the ten-week window that closed on 2026-03-31. On the business side, 12,000 firearms had been destroyed as of 2026-03-19 and more than 61,900 were claimed in a second window that closed on 2026-06-04. Whether the second figure includes the first is not stated anywhere opened here, so the two are reported separately and are not added together in this document.

A firearm is not a person, and the ratio between them is measured only on the group that complied. The 37,869 individuals who declared held about 1.8 firearms each. Nothing opened here reports what a household that did not declare is holding, and the error runs in both directions at once, since a person who owned a single prohibited rifle had the least to lose by handing it in while a collector had the most.

The licence figures cannot supply a ceiling either. Canada recorded 2,458,677 firearms licence holders in 2025 and 794,768 of those held the restricted class, both figures rising over the year before. The restricted class is the closest candidate for an upper bound, because many of the prohibited firearms had been registered in it before 2020. Many is not all, and no source opened here closes the membership of the group, so a count of restricted licences bounds nothing.

Needs a new measurementthe number of people. No source opened here counts the people who hold a prohibited firearm, either before the program or after it. Everything counted in this document is counted in firearms, and the population file therefore records the estimate as not derivable rather than assembling one out of parts that do not fit together.

Under what conditions does it arise?

1. The instrument was administrative, and that cuts both ways. A prohibition made by order took effect in twelve days without a parliamentary vote, and the same list has since grown from more than 1,500 models to more than 2,500 by the same route. That speed is now the question before the Supreme Court, and while the question is open the prohibition is in force and the shield over non-compliance is in force with it.

2. The population was never counted. A program cannot report a completion rate against a denominator it does not have. The federal estimate and the industry estimate differ by a factor of about 3.6 and the difference is the unregistered stock, which is by definition the part no registry can see. Every published figure for how well the program did depends on which of the two numbers the reader picks.

3. Participation is voluntary while the amnesty runs. Handing in a prohibited firearm is a decision each owner makes, and the consequence of not making it is deferred for as long as the amnesty is extended. A prohibition that carries no date carries no cost for waiting, and waiting is what the record shows.

4. The body that set the deadline is the body that keeps moving it. Three extensions and one replacement of the expiry rule have issued from the same cabinet that made the prohibition. Each one was a lawful administrative act and each one arrived shortly before the previous expiry, which means the pattern is visible only in the sequence and never in any single order.

5. Litigation now sets the clock. Since June 2026 the amnesty ends 90 days after four court files conclude, so the enforcement date is downstream of a schedule that neither the government nor the owners control. A challenge to a rule has become the thing that suspends the rule, without any court having ordered a suspension.

The condition that holds all of this in place is that nothing about the shortfall produces an event. No deadline passes, no report is due, and the prohibition on paper looks the same on a day when nothing is collected as on a day when a great deal is.

What has been tried?

Two directions have been tried and they are working against each other. One is administrative and has moved quickly, taking the prohibited list from more than 1,500 models to more than 2,500 and running the compensation program in phases. The other is judicial and has moved slowly, and it has now absorbed the deadline that the first direction depended on.

What was found?

Why is it still unsolved?

Enforcement absent — the prohibition took effect on the day it was made, and every mechanism that would have turned it into a change on the ground has been deferred by the same body that made it.

The first part is that the prohibition and the collection were separated in time, and only the separation has held. Declaring a firearm prohibited is an act of drafting and it was complete in twelve days. Getting it out of a house requires a program, a payment, a place to take it and a reason to go, and the individual half of that program did not open until nearly six years later and then stayed open for ten weeks. During those six years the prohibition was fully in force and almost entirely unenforced, and the amnesty is what made those two facts compatible.

The second part is that the deadline was always inside the same hands as the rule. An amnesty is an administrative act, so extending it costs nothing procedurally and produces no vote, no report and no finding of failure. Each of the four extensions arrived shortly before the previous one expired, and each was in itself unremarkable. The pattern only appears when the four are laid end to end, and no document in the sequence is required to lay them end to end.

The third part is the quietest and it is why the shortfall cannot be settled by argument. Nobody knows the denominator. If the stock is about 144,000 then just under half of it was declared by individuals, and if it is about 518,000 then about an eighth was. Both readings are available from federal documents, the second only as a figure the federal documents record without adopting, and neither can be tested because the difference between them is precisely the part that was never registered. A program that cannot state what fraction of its target it reached cannot be shown to have failed, which is also to say it cannot be shown to have worked.

What observation would mean it is solved?

Candidates — a. the amnesty expires on a stated date and is not extended again. b. an independent count of the prohibited stock is published and the collected share is reported against it. c. the Supreme Court rules and the government states what follows for the firearms still in private hands.

a alone is weaker than it looks. An amnesty can lapse without a single additional firearm being collected, and what would follow is a population of unknown size holding prohibited property with no shield. That is a change in legal exposure rather than a change in how many prohibited firearms are in circulation, and it could just as easily produce a fifth extension as an enforcement campaign. The expiry is also no longer a date at all, so this observation cannot even be scheduled until the court files conclude.

b is the strongest of the three and it does not exist yet. No count that reconciles the federal figure with the industry figure was found in this round. Until one exists, every statement about the shortfall is a ratio whose denominator the reader has to choose, which is why the two available choices differ by a factor of about 3.6. Building this measure is harder than it sounds, because the gap between the two estimates is made of firearms that were never registered anywhere, and it is possible that the true answer is that the count cannot be built rather than that nobody has tried.

c counts posture, not firearms. A ruling would settle whether the cabinet could act by order, and either outcome leaves the practical question untouched. If the prohibition stands, the same uncollected stock sits under the same program. If it falls, the program ends without the stock ever having been counted, and this document would close because the rule went away rather than because the rule was carried out.

The honest reading is that b has to be built before it can be read, a can be reported as a resolution while nothing has moved, and c is the one most likely to arrive first and to say the least about the thing this document is about.

What is it connected to?

Fills with researchthe closest structural relative in this corpus is the enforcement gap in cladding remediation in the United Kingdom, where buildings were officially judged unsafe and the volume of remediation actually started fell far below the volume judged. Both are cases in which a government made a formal determination and then did not deliver the removal or repair the determination implied. The mechanisms differ, because that one turns on who pays for work on a building and this one turns on voluntary surrender of private property under a shifting deadline, and relation type and evidence grade were not confirmed in this round. No relation to firearms policy in other countries was examined, and no adjacent Canadian program was pursued.

What these sources do not say

• How many firearms are actually in scope. The federal figure of about 144,000 and the industry figure of about 518,000 sit in the same documents without either being tested. No audit office, statistical agency or independent study opened here has tried to reconcile them. The sources also do not say which kind of gap this is: whether nobody has investigated the question, or whether firearms held outside any registry cannot be counted at all. Those two are different problems with different remedies and the material treats them as one.
• How many people hold prohibited firearms. Every count here is a count of firearms. No source converts them into owners, and none reports how many firearms a non-declaring household holds, which is the ratio the conversion would need.
• Whether the two business figures overlap. The 12,000 destroyed by 2026-03-19 and the more than 61,900 claimed by 2026-06-04 are reported separately and no source opened here states whether the second includes the first. The federal pages that would settle it refused automated retrieval.
• What the compliance percentages were divided by. A trade outlet reported a rate of 3.3 to 7.7 percent and a program cost of about 800 million Canadian dollars, or about 12,000 Canadian dollars per firearm collected. The article does not state the stock figures behind the percentages, so the range is not reproducible and is recorded here as that outlet arithmetic rather than as a finding.
• Which fiscal years the 67.2 million Canadian dollars covers. A lobby group asserts that this was spent before any firearm was bought back. No year is attached, and no federal accounting opened here confirms or contradicts it.
• When the Supreme Court will rule, or what a ruling would do to the program. Reporting places a decision in the following year and no source opened here analyses what either outcome would mean for firearms already declared, for payments already made, or for the amnesty that is now pegged to the conclusion of the files.
• What enforcement after the amnesty would look like. No source reports any prosecution, seizure or enforcement plan directed at undeclared prohibited firearms. This is an absence with an obvious reason, since the amnesty has not ended, but it means there is no evidence at all about the step the whole design depends on.
• Whether anyone who declared has been paid. The declaration totals are reported as declarations. Nothing opened here reports how many claims were assessed, how much was disbursed to individuals, or how long payment took.

See the evidence

No primary program record was read in full, and the evidentiary weakness of this document sits in one place that should be read plainly. The precise collection figures that this document is built on, being 68,717 firearms declared by 37,869 people and more than 61,900 claimed in the second business window, come from a private tracker rather than from a government publication. Every Public Safety Canada page that would confirm them directly, including the release announcing that more than 67,000 firearms were declared, refused automated retrieval and appears above as a row with no URL. So the federal confirmation of the federal result is present in this table only as an unopened row, which is why those findings are graded medium rather than high. Two federal documents did open and they carry the load elsewhere: the costing publication of 2021 supplies the cost range and the industry estimate, and the registered order of 2026-06-05 supplies the current amnesty rule word for word, which is the single most consequential fact here and the one held to the highest grade. The sourcing on the rest is one-sided in a way worth naming, since a trade publication, a firearms-owner lobby group and a private tracker are advocacy or trade sources with an interest in how this program is judged, and their figures are used here only where the direction of their interest does not change the arithmetic. Where sources overlap they agree, since the prohibition date and the initial scope appear in two opened sources and the amnesty extension of 2025 appears both in the tracker and in the registered order that amends it. Where they do not overlap the gap is left visible, so the relation between the 12,000 business firearms destroyed and the more than 61,900 later claimed is reported as unestablished rather than resolved by addition, and the compliance percentages published by one outlet are labelled as that outlet arithmetic rather than adopted. This is a research-based definition, so observation_refs is empty and provenance_mode is press-derived.

People affected

Estimated range Not derivable

The reason and what is missing are listed under “What is missing” below