Canada - Accessible AI Standard (CAN-ASC-6.2:2025)

Overview

CAN-ASC-6.2:2025 — Accessible and Equitable Artificial Intelligence Systems — is a National Standard of Canada developed by Accessibility Standards Canada and approved by the Standards Council of Canada. Published on December 3, 2025, and announced in a Government of Canada news release, the standard provides an equity-first, process-oriented framework for ensuring AI systems are usable by, and equitable for, people with disabilities. It addresses the full AI lifecycle — from planning and procurement to operation, monitoring, and continuous improvement — and is accompanied by accessible summaries and implementation materials. The standard is voluntary but may inform regulatory instruments under the Accessible Canada Act and is intended for broad use by public and private organisations, technology designers, procurers, researchers and regulators. See the official standard document at CAN-ASC-6.2 full HTML and the publication announcement at Canada releases world’s first standard on accessible and equitable AI.

Definitions

CAN-ASC-6.2 defines key terms to align interpretation across the AI lifecycle and accessibility contexts. Important definitions include "person with a disability" (as reflected in the Accessible Canada Act), "AI system" (systems performing tasks using algorithmic models, including automated decision-making systems and generative models), "accessible AI" (AI systems and supporting processes that are usable by people with disabilities), "equitable AI" (systems that produce fair outcomes and equitable benefits for people with disabilities), and "AI lifecycle" (planning, data collection, development, testing, deployment, monitoring and decommissioning). The standard also defines governance terms such as "technical committee" and "stakeholder participation" to ensure meaningful involvement of persons with disabilities at every stage. These definitions align with the standard’s cross-references to the Accessible Canada Act and international instruments referenced in the standard.

Governance and Institutional Framework

CAN-ASC-6.2 places governance at the centre of accessible and equitable AI. It requires organisations to embed accessibility and equity responsibilities into corporate governance structures, identify accountable owners for AI systems, and ensure decision-making bodies include or are accessible to persons with disabilities. The standard advises establishing accessible governance forums, accessible procurement policies, and oversight mechanisms that require documented risk assessments, equity impact analyses, privacy and data protections, and processes for receiving and responding to feedback from people with disabilities. Accessibility Standards Canada’s documentation also notes the voluntary standard’s relationship with the Accessible Canada Act and explains that standards developed by ASC "apply to federally regulated entities and can be recommended to the Minister responsible for the Accessible Canada Act"; organisations in federal domains should therefore consider alignment and potential regulatory uptake. The standard was created by a technical committee with a significant quota of lived-experience members and equity-deserving representation — reflecting ASC’s "nothing without us" principle — and ASC commits to review the standard within four years of publication to maintain institutional relevance. See the full standard for committee composition and legal context: CAN-ASC-6.2 full HTML and SCC notice: SCC: CAN-ASC-6.2.

Key Focus Areas

Clause structure concentrates requirements into three interlocking areas: Accessible AI (Clause 10), Equitable AI (Clause 11) and Organizational Processes (Clause 12), supported by education and annex material. Accessible AI requires that AI systems and the tools/processes used to create and manage them be usable by persons with disabilities (including data tools, programming environments, evaluation tools and consumer interfaces). Equitable AI requires organisations to ensure AI systems do not produce or amplify discriminatory outcomes for people with disabilities — mandating inclusive data practices, bias detection and mitigation, disaggregated performance reporting for disability groups, and real-world impact monitoring. Organizational processes mandate inclusion of persons with disabilities in governance and decision‑making, accessible procurement, accessible training and literacy, documented risk and harm assessments, and accessible feedback and remediation mechanisms. Education and training requirements ensure staff across roles understand accessibility and equity implications of AI, and that training materials and delivery are themselves accessible. Throughout, the standard emphasizes process-based, testable outcomes and points to annexes for practical examples and use cases to evaluate dataset fitness, testing approaches, and procurement checklists: see the summary and clauses at Summary of CAN-ASC-6.2.

Implementation Framework

CAN-ASC-6.2 is intentionally process- and outcomes-oriented to accommodate rapid AI evolution. The implementation framework encourages organisations to integrate accessible and equitable AI requirements into existing management systems (e.g., product development, procurement, privacy and risk management). Core steps include scoping AI use-cases to identify disability-relevant impacts; performing inclusive data audits to identify gaps and outliers; building inclusive labelling and annotation practices; conducting equity-focused validation and user testing with persons with disabilities; documenting accessible design decisions and fallback/alternative options; training staff and suppliers in accessible AI practices; and establishing continuous monitoring and feedback loops to capture real-world effects and trigger remediation. The standard anticipates supporting technical guidance and test methods will be developed in later parts to enable conformity assessment where needed. For access to the full clauses and annexes see the standard document at CAN-ASC-6.2 full HTML.

Monitoring and Evaluation

Monitoring under CAN-ASC-6.2 requires organisations to collect accessible, disaggregated metrics on AI performance for people with disabilities, to publish accessible summaries of system performance and known limitations, and to operate accessible feedback channels that persons with disabilities can use to report failures or harms. The standard requires periodic equity impact assessments, post-deployment audits, and retention of records to demonstrate compliance with internal policies and to inform continuous improvement. Organisations are also required to document mitigation steps and remediation timelines when harms or exclusions are identified. The standard’s monitoring approach is intended to be compatible with external conformity assessment and regulatory oversight if regulators choose to reference the standard in regulation or procurement.

Penalties, Liability, and Appeals

CAN-ASC-6.2 itself is a voluntary National Standard of Canada; ASC standards are not punitive instruments. However, Accessibility Standards Canada explains that its voluntary standards may be recommended to Ministers and can inform regulations under the Accessible Canada Act. Enforcement of obligations under the Accessible Canada Act (if regulations reference this standard or adopt its requirements) would fall to the Accessibility Commissioner and other federal authorities as appropriate. The Accessible Canada Regulations set out administrative monetary penalty ranges (minor, serious, very serious) and enforcement mechanisms administered by the Accessibility Commissioner, with penalty amounts varying by severity and scale of the regulated entity. For the regulatory penalty framework and guidance consult the Accessible Canada Regulations summary at Accessible Canada Regulations — summary and the ASC publication notice at Canada.ca news release. Organisations should treat the standard as best-practice guidance that may influence regulatory expectations and procurement criteria.

Relationship to Other Instruments

CAN-ASC-6.2 cross-references and is intended to align with the Accessible Canada Act, the Canadian Human Rights Act, the UN Convention on the Rights of Persons with Disabilities, and related Canadian and international standards (including EN 301 549, CSA ISO/IEC 42001 and other ICT accessibility standards). ASC states the standard will be reviewed within four years and will be supported by more precise technical guidance and annexes. The standard is positioned to complement existing AI ethics guidance and sectoral regulation while filling gaps related specifically to disability and accessibility. See ASC’s standard page for the list of referenced documents and normative references: CAN-ASC-6.2 full HTML.

International Alignment

ASC designed CAN-ASC-6.2 to align with international best practices and cited international standards such as CSA ISO/IEC 42001 and EN 301 549. The standard’s process-first, rights-based approach parallels international instruments (UN CRPD) and contributes a disability-focused complement to broader AI governance frameworks. Accessibility Standards Canada also coordinated technical committee membership with international experts and disability-rights representatives to ensure cross-jurisdictional relevance and to facilitate potential adoption or mutual recognition by other national standard bodies. For context on SCC accreditation and international positioning, see the Standards Council of Canada notice: SCC: CAN-ASC-6.2.

Implementation Timeline

Sources and References