What happened
The US is launching a new AI-powered risk assessment system to determine the conditions of prisoners' detention. The system will analyze data on convicts and predict the risk of violence. The pilot project will start in December 2026 in several male correctional facilities in the state of Michigan. The system will use software from the nonprofit technology company Recidiviz, which grew out of a Google project aimed at improving the criminal justice system through technology. The system will help determine whether prisoners can participate in state programs, change detention conditions, or be released.
The US is launching a new AI-powered risk assessment system to determine the conditions of prisoners' detention.
The system will analyze data on convicts and predict the risk of violence.
The pilot project will start in December 2026 in several male correctional facilities in the state of Michigan.
The system will use software from the nonprofit technology company Recidiviz, which grew out of a Google project aimed at improving the criminal justice system through technology.
The system will help determine whether prisoners can participate in state programs, change detention conditions, or be released.
Why it matters
The use of AI in the criminal justice system raises concerns about the accuracy of predictions and potential algorithmic biases. A previous ProPublica investigation revealed problems with a similar risk assessment system in the state of Louisiana, where the data on which the assessed prisoners could contain racial and socio-economic biases. In one case, the system classified a 70-year-old blind man who used a wheelchair as a person with a moderate risk of reoffending, despite having no disciplinary violations. Due to this assessment, the man automatically lost the right to parole.
The use of AI in the criminal justice system raises concerns about the accuracy of predictions and potential algorithmic biases.
A previous ProPublica investigation revealed problems with a similar risk assessment system in the state of Louisiana, where the data on which the assessed prisoners could contain racial and socio-economic biases.
In one case, the system classified a 70-year-old blind man who used a wheelchair as a person with a moderate risk of reoffending, despite having no disciplinary violations.
Due to this assessment, the man automatically lost the right to parole.
The experience of other states shows that errors can have direct consequences for prisoners' situations.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
A route planner searches possible journeys using explicit rules. What does this illustrate about AI?
What to watch next
The accuracy of the AI-powered risk assessment system and the potential for algorithmic biases will be closely watched. The system's ability to determine the conditions of prisoners' detention and predict the risk of violence will also be closely monitored.
The accuracy of the AI-powered risk assessment system and the potential for algorithmic biases will be closely watched.
The system's ability to determine the conditions of prisoners' detention and predict the risk of violence will also be closely monitored.
The impact of the system on prisoners' lives and the potential for errors will be closely watched.
The system's ability to reduce recidivism rates and improve public safety will also be closely monitored.
The potential for the system to be used in other states and countries will also be closely watched.