From the Public Safety edition of October 23, 2026
Law Commission of Ontario final report proposes provincial law for criminal-justice AI, with prohibited uses, a public registry and mandatory impact assessments
Law Commission of Ontario (Osgoode Hall Law School, York University) · Policing, courts and corrections / AI governance and law reform · Ontario, Canada (adaptable to other Canadian jurisdictions)
- Publisher
- Law Commission of Ontario
- Original publication
- Report dated October 2026; LCO publication page dated Oct. 22, 2026
- Source retrieved
- 2026-10-24
What happened
The LCO’s final report makes 29 recommendations for AI across Canada’s criminal justice lifecycle, after consultations with more than 1,400 participants. It recommends provincial legislation with a precautionary principle and four mandatory risk categories (prohibited, high risk, restricted use, low risk), and requires that criminal-justice AI systems be disclosed on a public registry, undergo a dedicated impact assessment and be audited on an ongoing basis. Its proposed prohibited list includes real-time remote biometric surveillance, individual offence-risk prediction including bail eligibility, emotion recognition, predictive policing and most agentic AI beyond low-risk back-office use. It also proposes an independent supervisory body, procedural rules for AI evidence and more defence and legal aid capacity. The report notes no Canadian government has yet enacted dedicated laws on police or court AI.
Why it matters
A full lifecycle framework that U.S. states, counties and courts can compare against their own patchwork of local ALPR ordinances, chambers orders and vendor policies. It names AI report drafting, ALPR, facial recognition and bail tools in one risk scheme.
Evidence and measured results
Opened the full report PDF and read the introduction, the legislative framework recommendation and Recommendation 12 on prohibited systems; publication date taken from the LCO publication page.
Limitations and uncertainty
Recommendations only; no government has adopted them. Ontario-focused and grounded in Canadian Charter law, so transfer to U.S. jurisdictions is limited. The prohibited list is described as a preliminary assessment.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-24; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Agencies and courts adopt AI tools under scattered local rules and face rising public challenge without a common risk framework.
- Stakeholders
- Police and court executives
- Ministry or attorney general policy staff
- Privacy commissioners
- Crown prosecutors and defence bar
- Procurement leads
- Discovery questions
- Can you list every AI feature in your evidence, records and camera systems?
- Which systems would fall in a prohibited or high-risk tier?
- Who would publish registry entries?
- Value hypothesis
- An inventory and tiering exercise could let an agency anticipate legislated duties and answer public questions with documented assessments.
- Potential engagement
- Six-to-ten week AI inventory and risk-tier mapping for one agency or court, ending in draft registry entries.
- Unsupported claims
- That Ontario or any U.S. state will enact these recommendations
- That tiering removes the need for local public process
- That vendor tools are compliant by default
Pre-sales engineering
Role takeaway
- Fit
- Justice agencies with several AI-enabled platforms and no central inventory.
- Architecture and integration
- Inventory spans evidence management, records, camera and analytics platforms; registry entries draw on vendor documentation and local configuration.
- Prerequisites
- Vendor disclosure of AI features
- Data-flow documentation
- Legal review of risk criteria
- Constraints
- Canadian Charter and Ontario law basis
- Prohibited list is preliminary
- Embedded AI features may be hard to identify
- Security
- Classify biometric and risk-prediction features separately; document data retention and sharing for each registered system.
- Proof of value
- Proposed: test whether every AI feature in one agency can be classified into the four tiers
- Proposed: draft registry entries for three systems and review with counsel
Delivery
Role takeaway
- Work
- Build the AI inventory, run impact assessments for high-risk systems and prepare registry entries.
- Dependencies
- Legal and privacy review
- Vendor documentation
- Executive sponsorship
- Ownership
- Police service or court administration, with any legislated supervisory body
- Skills and adoption
- Train project, legal and procurement staff to recognise embedded AI features and apply risk criteria.
- Governance checkpoints
- Inventory sign-off
- Impact assessment approval for each high-risk system
- Annual audit and registry update
- Acceptance criteria
- Proposed: all in-scope AI features inventoried and tiered
- Proposed: impact assessments completed before any new high-risk deployment
- Risks
- Recommendations may not be enacted
- Incomplete inventories from undisclosed vendor features
- Over-classification slowing low-risk uses
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
A public registry and ongoing audits presuppose an inventory of every AI component, including features embedded in evidence and records platforms.
Governance
Who approves, reviews and stays accountable for outcomes?
Proposes legislated risk tiers, a public registry, impact assessments and an independent supervisor; Recommendation 12 lists practices the LCO would prohibit.
Security and privacy
What data, permissions and controls need testing?
Treats biometric surveillance and biometric categorisation as prohibited or high risk; frames digital sovereignty as part of procurement.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Calls for added capacity for defence counsel and Legal Aid Ontario, including access to experts and forensic tools.
Procurement
What should contracts, pricing and exit terms secure?
Recommendation 18 ties AI pilots and procurement to digital sovereignty and the risk framework; vendors would face disclosure and assessment duties.
Operating model
Which teams own the service once it runs?
Would move police and courts from self-governance policies to statutory duties with external reporting and oversight.
Publication history
- 2026-10-23Public Safety · Issue 459 resources
Stable resource ID: law-commission-ontario-ai-criminal-justice-final-report-29-recommendations-20261022