Records Show Most CSIS Workplace Accidents Are Car Crashes During Surveillance

Government Files is The Canada Report's public-records analysis series examining government documents obtained through Canada's Access to Information (ATI) and provincial Freedom of Information (FOI) laws. These transparency laws allow members of the public to request internal government records from federal and provincial institutions. This article reviews documents released through those processes and summarizes what the records contain and what they show. While we strive for accuracy, this article represents an analysis and interpretation of the source material. For complete accuracy and full context, readers should review the original documents, which are available in full below.

Full Document

The complete document is available for download below:

For an organisation built around secrecy, the most common workplace hazard at the Canadian Security Intelligence Service turns out to be remarkably ordinary: the car. Records released under the Access to Information Act list more than 600 employee accident forms filed at CSIS between late 2021 and the summer of 2025, and the single largest legible category of incident is the motor vehicle accident — the occupational reality of an agency whose surveillance work is, in large part, a driving job.

The dataset was disclosed by CSIS in response to an access request (file 117-2025-135) seeking all Employee Accident Forms submitted by the agency's staff. It covers the period from 25 November 2021 to 1 August 2025 and runs to roughly 620 rows, each representing one reported incident. Employee names are withheld in full throughout the release, but the surrounding columns — date of injury, type of event, nature of injury, and the section or unit involved — are largely intact.

What the Documents Show

Each row records a single reported event, logged with the date it occurred, the date it was reported, the date it was passed to occupational health and safety, a short description of the type of event, the nature of any injury, and the section the employee worked in. Read together, the legible columns sketch a clear picture of what actually goes wrong on the job at Canada's domestic intelligence service.

Among the entries where the event type is readable, motor vehicle accidents are by far the most frequent — roughly three in five. The remainder is dominated by two other categories: "near missed" events, where something almost went wrong but didn't, and "threat to safety" incidents. A handful of entries are logged simply as "other," and one stands apart from everything else in the release: a single report classified as "chronic mental stress."

The injury column reinforces just how routine most of these forms are. The overwhelming majority are marked "No Injury," with a smaller share logged as "Minor Injury." Across the entire four-year span, the records contain no entry describing a serious or catastrophic injury in the legible fields — a notable detail for an agency whose work is popularly imagined as dangerous.

The section column is where the surveillance connection becomes explicit. The unit named most often, by a wide margin, is "Physical Surveillance" — the teams responsible for following and observing targets in the field. Other sections appear far less frequently and reflect the back-office side of the agency: Management Services, Human Resources, Corporate Services, Finance, Pay, Internal Security, Recruitment, and regional offices including the Calgary District, Edmonton District, Toronto Region, and the Manitoba & Saskatchewan Districts.

Surveillance Is, Mostly, a Driving Job

The dominance of motor vehicle accidents in the data lines up with the dominance of the Physical Surveillance section. Following a person of interest through a Canadian city means hours behind the wheel — in traffic, in parking lots, in winter conditions, often while trying to stay inconspicuous and keep eyes on a moving target. It is, statistically, where CSIS employees get hurt or come closest to it.

That reframes the day-to-day risk profile of the agency in a way the public rarely sees. The records suggest the typical CSIS safety incident is not a confrontation or an operational emergency but a fender-bender, a near-miss in a vehicle, or a low-speed collision during the long, unglamorous work of surveillance. The same field unit that has surfaced in The Canada Report's earlier reporting on the agency's BC region appears here as the part of CSIS most exposed to ordinary physical risk.

The steady presence of "threat to safety" entries adds a second dimension. Unlike the vehicle accidents, these are not described in the legible columns in enough detail to know what each one involved — whether an encounter in the field, a confrontation, or something else entirely. The forms record that the events were reported and routed to occupational health and safety, but the released columns do not explain their nature.

The One Chronic Mental Stress Entry

The lone "chronic mental stress" classification in the dataset is worth pausing on, precisely because it is alone. Workplace mental-health pressure at CSIS is not a hypothetical concern: it has been the subject of academic study and of The Canada Report's earlier reporting on how the agency's polygraph screening weighs on employees. Yet across roughly 620 accident forms spanning nearly four years, mental stress is formally logged exactly once.

The records do not explain that gap, and it would be speculation to read too much into a single field. But it raises a question the data cannot answer on its own: whether chronic stress at the agency is genuinely rare, or whether the Employee Accident Form — a tool built around discrete, datable physical events like collisions and falls — is simply the wrong instrument for capturing it. A cumulative, hard-to-pinpoint condition does not fit neatly into a form that asks for a single "date of injury."

What's Missing from the Data

The release is a log, not a narrative. Every employee name is redacted, so the records cannot be tied to individuals, and there is no free-text description of what happened in any given incident — only the categorical fields. That means the data shows that an event was a motor vehicle accident or a threat to safety, but never how serious it was beyond the broad "No Injury" / "Minor Injury" labels, who was at fault, or what the operational context was.

The records also offer no totals, no analysis, and no agency commentary. There is no indication of how these figures compare to other federal departments, whether the incident rate is rising or falling, or how CSIS responded to any individual report. The "reported to OHS" dates confirm that incidents were forwarded to occupational health and safety, but the outcome of that process is outside the scope of what was released. Anyone wanting to understand the agency's safety record in full would need the underlying incident reports, not just this summary log.

Why This Matters

CSIS rarely releases granular internal data, which is what makes a four-year run of accident forms unusual. On its own, the picture is reassuring in one sense — most incidents caused no injury — and revealing in another: the agency's frontline physical risk is concentrated in vehicles and in its surveillance teams, not in the dramatic scenarios that define the public imagination of spycraft.

It also fits a growing body of access-to-information disclosures that show the human and administrative side of Canada's intelligence service. The Canada Report has previously examined labour-code safety violations at the agency's Burnaby office and internal records describing a "toxic" culture in its BC surveillance unit. Taken alongside those files, this accident log adds a quieter but no less telling layer: the everyday occupational hazards of intelligence work, recorded one form at a time.

All figures referenced are drawn from Employee Accident Forms submitted by CSIS employees, released by the Canadian Security Intelligence Service under Access to Information request 117-2025-135, covering 25 November 2021 to 1 August 2025. The records consist of categorical incident logs with all employee names redacted; counts described here are based on the legible, non-redacted fields in the release.