Prohibits the department of corrections and community supervision from using artificial intelligence in evaluating the risk and needs principles used to measure rehabilitation of a person, in determining which incarcerated individuals may be released on parole or the level of supervision for individuals on parole; prohibits the department from using artificial intelligence when developing transitional accountability plans.
S10349 would amend New York’s executive law and correction law to bar the Department of Corrections and Community Supervision from using artificial intelligence in two core correctional functions: parole decision-making and the development of transitional accountability plans. In the parole context, the bill would prohibit the department from using AI, AI models, or AI systems to evaluate risk and needs factors, assess rehabilitation, determine whether an incarcerated person should be released to parole supervision, or set the level of supervision for a person on parole. It also requires that each release decision and supervision level be approved in writing by the Board of Parole or its designee.
The bill also revises the transitional accountability plan process, which is the individualized case management plan created when an incarcerated person enters DOCCS custody. Under the bill, those plans must still be comprehensive, dynamic, and tailored to programming and treatment needs, but the department would be barred from using AI in developing them. Each plan would also need written approval by the commissioner or the commissioner’s designee. The bill takes effect immediately.
If enacted, the bill would directly limit DOCCS’s use of automated decision tools in parole and reentry planning, while leaving the underlying statutory framework for parole risk-and-needs assessment and transitional accountability plans in place. It would add explicit statutory prohibitions on AI use and require written human approval for parole release, supervision levels, and transitional accountability plans. The affected parties would include incarcerated individuals, parole applicants, people on parole supervision, DOCCS staff, and the Board of Parole.
No committee transcript or vote history was provided, so there is no recorded debate or roll-call record to gauge support or opposition. Based on the bill’s text, the measure appears motivated by caution about automated decision-making in the correctional system and a preference for human review in parole and reentry planning. The available context does not show any formal sentiment from legislators, agencies, or stakeholders.
The main point of contention is likely to be whether artificial intelligence should be allowed to assist or influence parole and reentry decisions. Supporters would likely argue that these decisions are too consequential to rely on opaque or error-prone automated systems, and that individualized human judgment should remain central. Opponents may argue that AI tools can improve consistency, efficiency, and risk assessment, and that a categorical ban could limit the department’s ability to use modern analytic tools. The bill also raises practical questions about how broadly the AI prohibition would apply to decision support tools versus fully automated systems.