Ontario prison AI assigns black prisoners harsher living conditions (breachmedia.ca)

In Ontario, an AI tool known as the Security Assessment for Evaluating Risk (SAFER) has been found to assign harsher living conditions to Black prisoners compared to their counterparts. The algorithm, which operates by analyzing inmates' personal data such as arrest histories and disciplinary records to predict behavior, has disproportionately designated Black individuals to maximum security. A class action lawsuit highlights systemic racial biases inherent in the data that fuels SAFER, alleging that these discriminatory practices violate the Charter rights of Black inmates by denying them equal protection under the law. This situation raises significant concerns within the AI and machine learning community, particularly regarding the use of AI in sensitive contexts like the criminal justice system. Critics argue that relying on historical data fraught with racial bias perpetuates and amplifies existing inequalities. The Ontario Ministry of the Solicitor General has acknowledged the systemic discrimination faced by Indigenous and racialized individuals, yet no equivalent measures have been put in place for Black prisoners. As the SAFER algorithm continues to operate with minimal oversight and transparency, it serves as a cautionary tale about the risks of integrating AI into critical areas of public policy without addressing the ethical implications of biased training data.