Ontario’s jails are using a computer program to decide who goes to maximum security. That program, SAFER, is putting Black inmates in harsher conditions at rates that have nothing to do with their actual behavior. It is doing so disproportionately.
An investigation by The Breach and a new class action lawsuit expose the mechanics of this bias. The findings are damning. The province uses historical data to predict risk. That data is already skewed. The algorithm amplifies the skew.
How the SAFER algorithm works
The tool in question is the Security Assessment for Evaluating Risk, or SAFER. Ontario jails have used it since 2021. It assigns a risk score to every inmate. The score dictates security level. High risk equals maximum security. Low risk means less restrictive housing.
But how does it calculate risk? It looks at arrest records and disciplinary history. It does not look at guilt. It looks at past interactions with police. Those interactions are not random. They reflect systemic racism.
Journalist Desmond Cole examined the tool’s output. He dug into government data. He found a stark racial gap. Black people make up about five percent of Ontario’s population. Yet they accounted for nearly twenty-seven percent of maximum-security prisoners between 2022 and 5. That is not a rounding error. That is a pattern.
White people were underrepresented in that maximum-security category. Significantly so. The data shows that Black inmates are flagged as dangerous more often than white inmates with similar profiles. Why? Because their histories contain more police contacts. More charges. More arrests. Even when innocent. The tool treats arrest records as evidence of future behavior. It confuses contact with crime with criminality.
Why Ontario ignored Black prisoners
The class action lawsuit, filed by law firm Koskie Mizzly, claims the province knew this would happen. They had the data. They saw the disparity. Did they act? No.
Internal ministry training documents tell a different story than public statements. The documents admit that assessments like SAFER “would likely contribute to the overrepresentation” of Indigenous and racialized people. They knew the tool was biased.
So why not fix it for Black inmates?
They did fix it for Indigenous prisoners. The province adopted specific mitigation measures for Indigenous populations. They created safeguards. They recognized the unique history of colonialism and its impact on the justice system. But they did not extend similar protections to Black prisoners. They drew a line. Indigenous bias got a fix. Black bias got ignored.
This selective mitigation suggests a prioritization of optics over equity. It allows the system to claim it is addressing disparity while leaving the largest non-Indigenous minority to suffer the consequences.
Who designed this, and who is watching?
Grant Duwe is the researcher credited with designing SAFER. The Ministry of the Solicitor General oversees its use. Both were asked for comments. Both went silent. The Breach noted that neither responded to requests regarding the methodology or accuracy of the program.
This silence is telling. When a tool affects lives and liberties, and when that tool shows clear racial bias, transparency is mandatory. The refusal to explain raises questions about the validity of the risk scores themselves. If you cannot justify the inputs, you cannot trust the outputs.
The Ontario Ombudsman is watching now. The Ombudsman’s report, cited in the investigation, notes it has received complaints about SAFER’s impact. Specifically, complaints about the disproportionate effect on Black and Indigenous prisoners. An official ministry review of the program is currently underway.
Is there an alternative?
Some might argue that risk assessment is necessary. Jails must manage security. They must predict violence. But relying on historical arrest data is a lazy shortcut. It outsources moral judgment to code.
The current system assumes that past policing behavior is a valid proxy for future danger. It ignores the reality that marginalized communities are policed more heavily. It penalizes them for the sins of a flawed justice system.
- “Assessments like SAFER would likely contribute to overrepresentation” – Internal Ministry Documents
The lawsuit aims to hold the province accountable. It seeks to expose the negligence of ignoring known disparities. The ongoing review may change how the tool is used. It may tweak the weights. It may add new factors.
But changing a formula does not fix the root cause. The root cause is the justice system itself. It is the policing practices that generate the data in the first place. You can adjust the algorithm all you want. If the data remains poisoned by bias, the output will always be biased.
We are left with a question that has
