Canada wrote a daily four-hour out-of-cell minimum for Structured Intervention Units into law in 2019, but the advisory panel Public Safety Canada established found that in data through 2024 more than half of stays in each length band fell short of that minimum on at least three quarters of their days, and the comprehensive review the same statute requires was already overdue by 2024-07

What is happening?

In 2019 Canada abolished administrative and disciplinary segregation in federal penitentiaries and replaced it with a new form of confinement called the Structured Intervention Unit. The instrument was Bill C-83, which received royal assent on 2019-06-21 and came into force on 2019-11-30.

The replacement was not a bare renaming. Parliament attached a daily floor to it in the text of the statute. Section 36 of the Corrections and Conditional Release Act requires that a person held in such a unit be given at least 4 hours out of the cell each day between 7am and 10pm, and at least 2 hours of meaningful human contact within those hours.

The federal advisory panel that Public Safety Canada established to watch the rollout analyzed the administrative data of Correctional Service Canada and reported that the floor is routinely not reached. In its final report, received and announced by the Minister of Public Safety on 2025-01-27 and built on data running through 2024, the panel found that 58 percent of stays of 15 days or fewer, and 54.6 percent of stays of 16 days or more, failed to reach 4 hours out of cell on at least 75 percent of the days of the stay. On the contact standard the same report found that 55.2 percent of the shorter stays and 47.1 percent of the longer ones failed to reach 2 hours on more than half of the days. The panel concluded that the construction and operation of these units has not eliminated the experience of solitary confinement in Canada.

The statute also wrote its own correction into itself, and that clause did not fire either. Clause 40.1 of Bill C-83 requires that a comprehensive review of the Act be undertaken at the start of the fifth year after it came into force, by a committee of the Senate, of the House of Commons, or of both Houses. With a coming-into-force date of 2019-11-30, that start falls in 2023 or 2024. The annual report of the same advisory panel, published on 2024-07-15, records the review as already overdue.

Whose problem is this?

The body that operates the units is also the body that generates the record used to judge whether the units meet the standard. That arrangement is normal in correctional administration, and it is the reason the outside analysis of that record mattered so much.

Where does this problem end?

The boundary here is not whether the units are lawful but whether the daily minimums written into the statute are delivered and measured.

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

Now

What it should be

The target state in this document is not an outside proposal because the statute itself names the daily hours and the review deadline. Section 36 of the Corrections and Conditional Release Act sets the floor at 4 hours out of the cell and 2 hours of meaningful human contact on every day between 7am and 10pm. Clause 40.1 of Bill C-83 sets the deadline for a comprehensive review of the Act at the start of the fifth year after it came into force. Both numbers belong to Parliament, and neither had to be supplied from anywhere else.

How big is it?

What the sources do give is the shape of the exposure.

Volume. In 2023 there were 433 stays of 61 days or more, a rate of 33.7 per 1,000 inmates. The panel estimated 534 such stays in the first half of 2024, a rate of 38.0 per 1,000, and recorded that this is close to the level seen in the final year of the segregation regime the units replaced.

Duration. Across 2020-04 to 2023-03 the internal audit of Correctional Service Canada found a median stay of 21 days overall, 26 days for Indigenous inmates and 6 days for women.

Distribution. The daily minimum in Section 36 is the same for every person held in one of these units and does not vary by group. Measured against that single standard, the panel reported on data through 2024 that 25.9 percent of Black inmates, 23.6 percent of Indigenous inmates and 19.3 percent of white inmates held in these units stayed longer than 60 days, and that 37.5 percent of inmates whose mental health was recorded as deteriorating stayed longer than 62 days.

Needs a new measurementthe headcount: no source opened here gives the federal in-custody population, which is the denominator that the transfer rate of about 1 percent and the rates per 1,000 inmates are expressed against. The departmental summary page opened here links onward to a fuller report rather than carrying the figure. Without it these rates cannot be turned into a number of people, and the counts below are counts of stays rather than of distinct individuals, because one person can have more than one stay in a year.

Under what conditions does it arise?

1. A minimum that is written as an hourly entitlement can be missed on a single day without any step in the system responding to it. The statute states the hours. The material opened here describes no automatic consequence attached to a day on which those hours are not delivered, and the shortfall becomes visible only when someone aggregates the daily records afterward.

2. The correction the statute does provide runs through a discretionary decision. When a stay grows long, an independent external decision-maker may order the person released from the unit. Between 2022 and 2024-08 the panel found that 89 percent of those decisions authorized continued confinement, and that the rate at which release orders were issued varied widely among the decision-makers.

3. An order to release is not the same as a release. Of 41 release orders in that window, 44 percent were followed by actual release within 30 to 40 days, and 37 percent of the people covered by such an order were still in a unit more than 60 days later. For people ordered to remain, the comparable share past 60 days was 3 percent.

4. The record is kept as offers rather than as delivered hours. The internal audit records that between 68 and 77 percent of proposed out-of-cell activities were declined. The panel took issue with framing compliance as offers made, on the ground that what Section 36 requires is time actually spent out of the cell.

5. The statutory review has no named starter inside the operating institution. Clause 40.1 assigns the review to a committee of the Senate, of the House of Commons, or of both Houses. A commentary published on 2025-05-27 by a former independent external decision-maker states that the review had not started and that the statute does not clearly say who is responsible for beginning it.

What has been tried?

Two of these ran for years and produced measurement rather than change. The panel measured the same shortfall in report after report, and the internal audit measured it again from inside, and the daily hours in the data did not move to the level the statute names.

What was found?

Why is it still unsolved?

Enforcement absent — the standard exists in the statute, it is measured, the measurement shows it is missed, and nothing in the design turns that measurement into a consequence.

An entitlement that is measured by the same service that delivers it produces a record that nobody outside that service is required to check. That is not a criticism of the record, which is what the panel used and what the internal audit used. It is a statement about where the checking lives. For five years the checking lived in a panel with an end date, and that end date has passed.

The second part is that the escape hatch the statute does provide is discretionary at both ends. An independent external decision-maker may order a person out of a unit, and between 2022 and 2024-08 that instrument authorized continued confinement in 89 percent of decisions. When an order to release did issue, it was followed by release within 30 to 40 days in 18 of 41 cases. A remedy that is discretionary to grant and slow to execute does not function as a backstop for a daily entitlement, because the entitlement is counted in hours and the remedy is counted in weeks.

The third part is that the statute contains a review clause and the clause is silent about who starts the clock. Clause 40.1 names a committee of the Senate, of the House of Commons, or of both Houses, in the passive form of a review that shall be undertaken. The panel recorded the review as already overdue in 2024-07. The panel that recorded it ceased to exist on 2024-12-31. A deadline with no named holder produces no event when it passes, and a missed deadline that produces no event produces no record of having been missed.

What observation would mean it is solved?

Candidates — (a) the daily hours actually delivered reach the statutory floor for the large majority of stays, measured across a full year (b) a committee of the Senate, of the House of Commons, or of both Houses undertakes the comprehensive review that Clause 40.1 requires and reports (c) release orders from independent external decision-makers are executed within days rather than weeks, and the share of stays past 60 days falls.

(a) has a measurement problem that is now structural. The daily hours are recorded by the service that operates the units, and the body that analyzed those records and published the analysis ended its mandate on 2024-12-31. No source opened here names a successor doing the same work. Until one exists, a future report of improved compliance and a future absence of independent analysis are not distinguishable from outside, because both look like silence.

(b) is the only candidate with a fixed legal referent, and it is also the narrowest. A review can be undertaken and can report without changing a single hour in the daily schedule of anyone held in one of these units. What it would produce is a public record of the gap that currently has no owner, which is a precondition rather than a result.

(c) is the most direct test of whether the remedy works, and it is the one with the smallest denominator. Forty-one orders across roughly two and a half years is a thin base to read a trend from. It would take several years of similar data to distinguish a change in practice from ordinary variation in a small number.

What is it connected to?

Fills with researchhealth care in federal custody, the composition of the federal inmate population, segregation practice in provincial and territorial custody under separate statutes, separation of young persons under the youth justice framework, and the general question of what happens to statutory review clauses in other federal acts. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• The federal in-custody population. No source opened here gives the count that would serve as the denominator for the transfer rate of about 1 percent or for the rates expressed per 1,000 inmates. The departmental summary page opened here links onward to a fuller report rather than carrying the figure, so the volume figures in this document stay as counts of stays and as rates.
• Whether the statutory review has begun since 2025-05. The most recent source opened here on that question is a commentary dated 2025-05-27 stating that it had not started. Nothing opened here covers late 2025 or 2026, so the status on 2026-09-08 is unknown rather than confirmed either way.
• What analysis replaced the panel after 2024-12-31. No source opened here names a body that has since taken up the same analysis of the administrative data and published it, and none states that no such body exists.
• Why proposed out-of-cell time was declined. The internal audit records that between 68 and 77 percent of proposed activities were declined and the panel took issue with counting offers rather than delivered hours, but no source opened here quantifies the reasons for declining, whether they concern what was offered, safety, scheduling, or something else. The quantitative basis for the criticism by the panel was not visible in the material opened here.
• A response addressed to the compliance percentages themselves. The published positions opened here are the operational announcement of 2022-10-29, the ministerial release of 2025-01-27 receiving the final report, and the internal audit published 2025-06-08 with management responses attached. No source opened here carries a reply from Correctional Service Canada addressed specifically to the finding that more than half of stays in each length band missed the 4-hour minimum on at least three quarters of their days.
• Whether current constitutional litigation targets this regime. Challenges to the pre-2019 segregation provisions ended when those provisions were repealed in 2019. This round searched for a current case aimed at the units themselves and could not identify a case name, court or status from a primary source. Nothing opened here asserts that such a case exists, and nothing opened here asserts that none does.
• Whether the 2022 operational changes moved the daily hours. The tracking system and the additional counsellors were announced in 2022-10-29 and the panel continued to report shortfalls afterward, but no source opened here isolates what those measures changed.

See the evidence

No primary correctional record was read. The two statutory texts were opened directly and are the only primary documents here read as originals, and they supply the daily minimums and the review deadline that this document measures against. Everything about performance comes from two public bodies analyzing the administrative data of Correctional Service Canada — the advisory panel established by Public Safety Canada, and the internal audit and evaluation function of the service itself — and both are secondary to the underlying records. Where the two agree, they agree on direction rather than on the same measure: the panel counts days that fall short of the statutory hours, and the audit counts activities that were proposed and declined, which is why the panel objects to that second frame as a compliance measure. Where the record simply stops, it is left visible rather than filled: the federal in-custody population is missing, the status of the statutory review after 2025-05 is unknown, and no successor to the analysis the panel performed is named in any source opened here. This is a Path A output, so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range Not derivable

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