Wonders & Oddities | Artificial Intelligence

Michigan Prisons to Implement AI for Inmate Risk Assessment

2 November 2026 2 min read هدهد تك | Hudhud Tech

Michigan is set to implement a new AI system across its prisons to classify incarcerated individuals based on their perceived danger levels.

According to Detroit’s Fox2, the Michigan Department of Corrections (MDOC) is moving forward with an inmate categorization algorithm designed to sort individuals into various housing levels.

In Michigan, incarcerated individuals are divided into five housing levels according to their security classification. Previously, these classifications were determined by a committee of staff appointed by the Warden at each facility. The state now plans to rely heavily on a non-profit tech company called Recidiviz for this decision-making process.

Recidiviz originated as a data organization initiative at Google, aimed at standardizing criminal justice data and enhancing the system nationwide. Its AI-integrated software platform is designed to identify individuals eligible for state programs, facilitate reclassification or release, and coordinate access to resources like transitional housing and employment.

In a press release, MDOC stated that the Recidiviz software will be used to “optimize placement of incarcerated individuals,” starting with a pilot program in several men’s facilities in December. The agency claims the AI tool is expected to improve its ability to predict the risk of future violence among incarcerated individuals by approximately 30 percent, although the basis for this figure remains unclear.

Housing designations are crucial for incarcerated individuals, as they determine eligibility for programs and the type of population they will be housed with during their sentences.

Despite the state’s rationale, the partnership between MDOC and Recidiviz raises significant questions. The type of data collected will greatly influence the accuracy of the AI system. A ProPublica investigation into a similar program in Louisiana revealed that the data used contained inherent racial and socioeconomic biases.

This bias has led to a more punitive carceral system, adversely affecting those striving for rehabilitation. For instance, a 70-year-old individual who is blind and wheelchair-bound was classified as a “moderate risk” of reoffending despite having a clean disciplinary record, resulting in an automatic loss of parole eligibility.

While Michigan’s system focuses on housing assignments rather than parole decisions, it is unlikely that software alone will resolve the state’s ongoing corrections crisis. If Louisiana’s experience with AI is any indication, it could exacerbate existing issues.

The Michigan Department of Corrections was contacted for comment.

More on incarceration: What Are the Ethics of Strapping VR Headsets on Inmates in Solitary Confinement?