London Police Adopts AI Governance Framework

The London Police Services Board approved a new artificial intelligence governance policy that formalizes how the London Police Service will evaluate, approve, deploy, and report on AI tools. The framework requires that AI remain subject to meaningful human oversight and be justified, proportionate, and consistent with legal and ethical obligations. The board will receive an annual AI Technology Compliance and Risk Report, portions of which may be summarized publicly when operational sensitivities exist.

The policy creates tiered controls and explicit governance steps for AI tools. Key operational elements are:

• •Board approval for moderate- and high-risk technologies, defined by visibility, potential privacy impact, and human-rights implications
• •Requirement that AI use be demonstrably lawful and further policing purposes only when benefits outweigh risks
• •Annual AI Technology Compliance and Risk Report with a public-facing summary for non-sensitive content
• •Mandates for the chief to develop procedures, training, monitoring, performance evaluation, and mitigation plans within 12 months of any new tool approval

The policy explicitly cross-references compliance with the Canadian Charter of Rights and Freedoms, human-rights legislation, privacy statutes, and policing laws. The board framed AI as an efficiency opportunity with attendant risks: "AI technologies are becoming increasingly embedded in policing," said board chair Ryan Gauss.

London's framework mirrors governance models already adopted in Toronto, York, and Peel jurisdictions, signaling growing municipal-level standardization in the absence of a provincial framework. Practitioners should note two practical consequences. First, procurement and vendor evaluations will now need to include documented risk assessments, human-in-the-loop designs, and auditability provisions to satisfy board review. Second, operational pilots such as using AI to draft reports from body-worn cameras will be subject to formal oversight, training requirements, and post-deployment audits. That use case, described by local reporting as a forthcoming efficiency for drafting officer reports, raises tangible questions about transcription accuracy, hallucination risk, redaction and retention policies, and chain-of-evidence integrity.

From a governance design perspective, London's approach emphasizes accountability, proportionality, and transparency rather than outright prohibitions. That makes the policy actionable for product teams: expect to provide compliance artifacts, privacy impact assessments, testing results, and mechanisms for human review and appeal when seeking adoption.

Municipal adoption creates precedents for procurement language and audit requirements; vendors and in-house teams should prepare standardized risk assessment templates, logging and retention controls, and operator training curricula. Provincial coordination remains absent, so divergence across services could persist until Ontario issues a centralized framework.