Black prisoners in Ontario’s jails are being assigned to harsher living conditions than other prisoners, through the use of an artificial intelligence (AI) tool that claims to predict their behaviour. A class action lawsuit says the province was aware its use of the software could disproportionately target Black prisoners, but went ahead with it anyway.
The Security Assessment for Evaluating Risk (SAFER) program has been operating quietly in Ontario’s jails since early 2021. SAFER inputs a prisoner’s personal information—including arrests, charges, and disciplinary records—into an algorithm. The program assigns each prisoner a score from 0 to 100 that determines whether they’ll be placed in minimum, medium, or maximum security detention.
Critics of the program argue that the data that SAFER is fed is racially biased: they cite documented patterns of police and courts handing out more severe punishments to Black people because of anti-Black racism. SAFER then uses that data to make harsher risk assessments of Black people who are sent to jail.
The ministry responsible for Ontario’s prisoners agrees. It wrote in internal training documents viewed exclusively by The Breach that “Indigenous and racialized individuals face systemic discrimination in our justice system … As a result, assessments like SAFER would likely contribute to the overrepresentation of Indigenous inmates in maximum security.”
Despite this, the ministry has been using SAFER for five years. And while the province has included several measures in its rollout of SAFER to reduce the number of Indigenous prisoners in maximum security, it does not appear to have taken such steps for Black prisoners. “We are continuously evaluating to determine if it is necessary to make similar adjustments for other groups,” the same training document from the Ministry of the Solicitor General says.
Nana Yanful, a human rights lawyer with a focus on racial justice, slammed the government’s decision not to curb the number of Black prisoners in maximum security, as it has done for Indigenous prisoners. “They are knowingly contributing to a program that will further exacerbate the conditions of confinement for Black folks,” Yanful said in an interview. “They just don’t care that this is going to impact them more negatively than non-Black incarcerated people.”
Safer for whom?
Shauna first learned about SAFER in 2024 while trying to schedule a visit with her partner, who is incarcerated in an Ontario jail. (The Breach has changed Shauna’s name and withheld the name of her partner’s jail, as she fears he may face retribution if identified.) “When I was inquiring about how to visit, I was told that I would have to wait to find out his score,” Shauna told The Breach.
“That’s when I got a phone call from him. He told me they run some program in there—he didn’t know the name of it—where they assign you a security level.” Shauna’s partner, who is Black, said her ability to visit him in-person would depend on his SAFER score. “If he’s in max, it would be one video visit a week. If he was in medium, I believe it was two in-person behind-glass visits.”
Shauna said her partner learned about SAFER from other prisoners. When The Breach asked if he had inquired about his SAFER score, she said, “He’s been requesting it for about a month now, and he keeps being told, ‘Yeah, they’ll bring it.’ He’s even requested it from sergeants, and nobody has come to him with it yet.”
Security designations determine what level of access a prisoner will have to visits, programming, and living spaces. The class action lawsuit describes the conditions in medium- and maximum-security spaces: “more restrictive cells and environments, with less access to movement, activities, programs, and amenities.” It says prisoners’ freedoms are severely limited in these settings.
The class action lawsuit, filed in 2025 by Koskie Minsky LLP, is based on the unequal detention outcomes the SAFER assessments have produced. Black people are the most likely group to be placed in maximum security. While Black people make up only 5.4 per cent of Ontario’s population, they represent nearly 27 per cent of all prisoners housed in maximum security detention from 2022 to 2025, according to government data analyzed by University of Toronto criminologist and professor Scot Wortley. By contrast, white people comprise 63.3 per cent of Ontarians, but only about 41 per cent of prisoners held in maximum security are white.
Black women are particularly affected: in a 16 month period between 2024 and 2025, over twice the proportion of Black women received a maximum-security designation from SAFER, compared to white women.
The class action lawsuit claims that SAFER violates Black prisoners’ Charter rights by denying them access to equal protection and equal benefit from the law. “Ontario could have reduced the discriminatory impact on Black inmates by adopting similar mitigation measures as have been adopted for Indigenous and women inmates,” the lawsuit’s statement of claim says. “Ontario failed to adopt such measures.”
Caitlin Leach, a lawyer with Koskie Minsky who is working on the class action, says arbitrary decisions in the criminal justice system—which includes choices made by police, judges, protectors, and jail guards—can contribute to bias in the SAFER data. Leach offered as one example “a police officer’s decision to charge someone with a particular offense that becomes part of their criminal record.”
“There might be a Black person who’s charged with an offense, and a white person in the same circumstances might not be charged with that offense. So when they are incarcerated, they have a different kind of criminal record that’s then being used in the algorithm.” Leach emphasized that the ministry has not disclosed exactly how SAFER works. The focus of the lawsuit is on the program’s disproportionate outcomes.
Research from the Ontario Human Rights Commission analyzing Toronto Police Service data from 2013 to 2017 found that “despite being charged at a disproportionately higher rate, Black people were overrepresented in cases that resulted in withdrawn charges. Their cases were also less likely to result in a conviction compared to cases involving White people.”
Four out of five people in Ontario jails are being held on remand—meaning they’ve been charged but not convicted. It’s unclear whether SAFER penalizes prisoners for having serious charges, even when they have not been convicted of those charges.
Lindsay Jennings, a researcher with the Tracking (In)Justice Project, also pointed to the arbitrary nature of discipline by jail staff. “It’s all subjective: who you’re talking to that day, if they’re in a bad mood, if they like you as a prisoner…or they know that you’re connected to advocacy on the outside, you’re gonna get treated very differently.”
Jail staff can override the SAFER program’s decisions and assign prisoners to a different security classification. Wortley’s analysis of limited override data showed that 39 per cent of overrides involved Indigenous prisoners, and 36 per cent involved Black prisoners. However, Wortley’s report notes that the data does not show whether the override assigned the prisoner to a higher or lower-security facility than the SAFER risk assessment recommended.
Ministry documents describe strategies to address systemic discrimination against Indigenous people. For example, jail staff must automatically review the score of an Indigenous prisoner when it falls within the lowest ranks of a maximum security placement. Staff can then choose to override the decision and assign the prisoner to a lower security level. During intake and release, Indigenous prisoners also work with Native Inmate Liaison Officers, who consider the intergenerational impacts of colonialism on Indigenous people.
Yanful, the human rights lawyer, highlighted the lack of information about the use of overrides for Black prisoners. Ministry training documents allow staff to override SAFER decisions to assign a higher security level if they believe a prisoner is a member of a gang. “I’ve had clients before who say, ‘I literally am not part of this gang that the police have alleged I’m part of, I just live in this neighborhood where this group of people is alleged to be operating,’” she said.
Yanful questioned the criteria jail staff are using to determine a prisoner’s alleged gang affiliation. “Discretion is often used against Black folks to their detriment,” said Yanful. “So I would approach anything with these overrides, particularly around gang affiliation that’s so coded with race, with a lot of scrutiny.”
Turning human bias into algorithmic bias
Very little is known about the SAFER algorithm, and the exact data it’s fed to assess prisoners. The ministry describes SAFER as “an automated, predictive tool for evaluating an inmate’s security risk level.” It does not explain its claim that the program can predict the future behaviour of prisoners.
The ministry document says SAFER was created by an external researcher using 10 years of historical prisoner data, including criminal charges and misconduct reports. The ministry says the researcher “identified the predictive factors for violent and/or frequent misconducts among inmates.” The document claims, without providing evidence, that SAFER’s ability to predict these incidents is as accurate for Indigenous and racialized prisoners as it is for other groups.
The Breach contacted the Ministry of the Solicitor General with questions about SAFER. We asked for evidence of SAFER’s alleged ability to make accurate predictions; we also requested evidence for ministry claims that SAFER has led to a “reduction in violence and occupational stress injuries.” The ministry did not respond to multiple written requests for comment.
Ministry documents identify Dr. Grant Duwe, research director at the Minnesota Department of Corrections, as the designer of the program. The Breach reached out to Duwe to ask about SAFER’s algorithm, inputs, and accuracy. Duwe, too, did not respond to multiple requests for comment.
A recent report by Ontario’s Ombudsman cites 126 complaints from prisoners about SAFER. According to some complaints, staff have told prisoners they can reduce their SAFER scores by participating in jail programming, but information about which programs lead to reduced scores is not available. The report notes that the Ombudsman “also heard concerns about the disproportionate impact of this tool on Black and Indigenous inmates.”
The emerging use of AI in Canada’s jails and prisons comes with serious threats to accuracy and accountability, according to a paper by the Law Commission of Ontario. The paper explains why such tools might initially be attractive to policymakers and the public: “The simplistic perspective here is that if AI algorithms could rightly assess risks posed by individuals in the criminal justice system, such tools would eliminate human bias.”
However, the paper warns about “the potential for spurious correlations the justice system should not rely on,” like using a person’s race or postal code to guess their likelihood of recidivism or relying on justice system data that’s long been riddled with bias. The paper cautions that AI operates as a “‘black box’ … incapable of rationalizing or explaining the decisions or recommendations it makes.”
The Ombudsman’s report notes that a ministry review of the SAFER rollout is currently underway. But that rollout began quietly in 2021, and the program has now been implemented in most jails across Ontario. Class actions typically take years, and while the lawsuit creeps forward, SAFER may continue operating unchanged—with much still untold about its impacts.
“We need media that enlarges the sense of what’s possible.”
Naomi Klein, journalist and author
Want to support hope-fuelled journalism? Become a sustaining member.