How to Implement CAN-ASC-6.2:2025 Accessibility Requirements for AI Systems?

Artificial intelligence is transforming how organizations operate, make decisions, and serve their users. Yet as AI systems become increasingly integrated into our daily lives, a critical question emerges. Are these intelligent systems truly accessible and equitable for people with disabilities?

Canada's new accessibility standard for AI, CAN-ASC-6.2:2025, provides a comprehensive framework. It ensures that artificial intelligence systems benefit everyone, including people with disabilities.

What is CAN-ASC-6.2:2025?

CAN-ASC-6.2:2025 is Canada's first accessibility standard specifically designed for artificial intelligence systems. Developed by Accessibility Standards Canada and published in December 2025, this groundbreaking standard represents a significant advancement in AI accessibility compliance and equitable AI development. It aligns with existing standards and regulations, including

• The Accessible Canada Act
• The United Nations Convention on the Rights of Persons with Disabilities and
• The Canadian Human Rights Act

The standard establishes concrete requirements for organizations that develop or procure artificial intelligence applications. Whether you're building AI tools, purchasing AI solutions or implementing AI-powered services, CAN-ASC-6.2:2025 provides the accessibility requirements and equitable AI guidelines you need to follow.

Goals and Purpose of CAN-ASC-6.2:2025

CAN-ASC-6.2:2025 establishes a clear vision for what accessible and equitable artificial intelligence systems should achieve.

• Give equitable benefits to people with disabilities: Artificial intelligence systems must deliver the same advantages and opportunities to people with disabilities as they do to other users
• Avoid causing unfair harm: AI systems must be designed and monitored to prevent discrimination, privacy violations, exclusion and loss of autonomy, with particular attention to people with disabilities
• Protect rights and freedoms of people with disabilities: AI must never be deployed in ways that undermine the dignity, privacy or independence of people with disabilities
• Treat people with disabilities with respect and give them choices: People with disabilities must have meaningful participation in AI decisions and retain the option to decline AI systems in favor of human alternatives

These goals reflect a human-rights-based approach to artificial intelligence governance and compliance with accessibility standards.

The Broader Context: Aligning with Other Standards

CAN-ASC-6.2:2025 doesn't exist in isolation. It aligns with several important standards and regulations:

• CAN-ASC-EN 301 549:2024: The broader standard for accessibility requirements in ICT products and services
• CSA ISO/IEC 42001:25: The international standard for AI management systems
• CSA ISO/IEC 30071-1-20: The international code of practice for creating accessible ICT products
• CSA ISO/IEC 29138-1:19: International guidance on user accessibility needs

Understanding these connections helps organizations take a comprehensive approach to accessibility and equitable artificial intelligence.

Scope of CAN-ASC-6.2:2025

The first nine clauses of CAN-ASC-6.2:2025 cover foundational elements, including the Standard Development Organization (SDO) responsible for maintaining the standard, the technical committee that developed it, legal obligations organizations must fulfill and guidance on how the standard should be implemented and interpreted.

Clauses 10 through 13 contain the substantive requirements for organizations developing, deploying, procuring or managing artificial intelligence systems. These requirements address:

• Accessible AI (Clause 10): Ensuring that artificial intelligence systems and tools are usable by people with disabilities
• Equitable AI (Clause 11): Guaranteeing that artificial intelligence applications treat people with disabilities fairly
• Organization processes to support accessible and equitable AI (Clause 12): Enabling organizations to build and deploy artificial intelligence in equitable and accessible ways
• Education and training (Clause 13): Teaching people about artificial intelligence and accessibility to make artificial intelligence systems more inclusive

This comprehensive scope ensures that accessibility and equity are woven throughout organizational processes and not treated as an afterthought or add-on to AI systems.

The Four Core Pillars of CAN-ASC-6.2:2025

The standard is structured to provide comprehensive guidance across four key dimensions of accessible and equitable artificial intelligence systems.

1. Accessible AI

Accessible AI requires that people with disabilities can fully participate in every step of the artificial intelligence lifecycle. This includes:

1. Design and development: People with disabilities should be involved in designing and contributing to the technical implementation of accessible and equitable artificial intelligence systems
2. Testing and quality assurance: Accessibility testing with people with disabilities is essential for identifying barriers
3. Procurement: Organizations must include disability perspectives when procuring AI solutions and evaluating vendors for accessibility and equity compliance
4. Implementation and deployment: Users with disabilities must be able to meaningfully use and benefit from the final AI product
5. Monitoring and reviewing: People with disabilities should participate in ongoing assessment, auditing and continuous improvement processes to ensure AI systems maintain accessibility and equity standards over time

Organizations must ensure that all tools used to build and manage AI meet accessibility standards. This includes making the AI systems themselves accessible, as well as ensuring that outputs from artificial intelligence systems are compatible with assistive technologies.

When artificial intelligence affects workers, customers or the public, organizations must provide clear, plain language information about how AI systems make decisions and allow people to challenge those decisions. Additionally, organizations must offer accessible feedback mechanisms so people with disabilities can voice concerns about AI accessibility.

A practical example of accessible AI in action: If an organization uses AI for sign language interpretation, deaf users must retain the choice to request a human interpreter instead, particularly in critical contexts like healthcare or legal proceedings.

2. Equitable AI

Equitable AI ensures that artificial intelligence systems treat people with disabilities fairly and don't perpetuate discrimination or cause harm. Key requirements include:

1. Equitable Access to AI Benefits: Organizations must take deliberate steps to ensure people with disabilities benefit equally from artificial intelligence systems. This means including people with disabilities in training data, testing AI tools with diverse user groups and sharing performance metrics separated by disability status. Organizations should continuously track real-world effects and use feedback to improve artificial intelligence systems over time.
2. Preventing Harm from AI Systems: Harm from artificial intelligence can take many forms. Discrimination, privacy violations, exclusion or loss of autonomy to name some. CAN-ASC-6.2:2025 requires organizations to proactively identify and minimize these harms. Specifically, organizations must:
   • Focus on risks that impacts people with disabilities
   • Address cumulative harms that build up over time
   • Use equitable methods to assess accuracy across different user populations
   • Protect personal and disability-related information
   • Avoid unfair treatment of persons with disabilities
   • Ensure informed consent before deploying AI systems
   • Avoid perpetuating harmful stereotypes about people with disabilities
3. Respecting Rights, Dignity and Choice: Artificial intelligence must never be weaponized to violate the rights and freedoms of people with disabilities. Organizations must avoid using AI to inequitably monitor or track people with disabilities and must not use AI systems to judge people based on physical characteristics, movement patterns, facial expressions or emotional states. People with disabilities must maintain the ability to make informed choices and be treated with respect. To achieve this, organizations must:
   • Include people with disabilities in all steps through meaningful participation throughout AI design, implementation and governance
   • Share clear and accessible information that is transparent and easy-to-understand explanations of how AI systems work, what data they use and what decisions they make
   • Offer meaningful choices like the option to choose human alternatives instead of AI
   • Avoid misleading or manipulating people and communicate honestly about AI capabilities, limitations and impacts
   • Establish clear accountability within the organization for AI decisions and their impact on people with disabilities
4. Supporting Equitable and Accessible AI Research: Organizations that support AI research must also support research that helps make AI equitable and accessible for people with disabilities.

3. Organizational Processes to Support Accessible and Equitable AI

Moving beyond high-level principles, CAN-ASC-6.2:2025 specifies detailed organizational processes and practices that support the development and deployment of accessible and equitable artificial intelligence systems.

1. AI Governance with Disability Inclusion: Organizations must establish equitable AI governance structures that include people with disabilities in decision-making roles. This means hiring people with disabilities as employees, board members and advisors to ensure disability perspectives inform AI strategy and oversight.
2. Planning and Risk Assessment: Before deploying any artificial intelligence system, organizations must create a clear plan that includes human decision-making options and emergency protocols. Risk assessments must consider both major and minor harms, with particular attention to impacts on people with disabilities. Organizations must use equitable methods to evaluate risks and accuracy.
3. Informing the Public About AI Use: Transparency is essential for equitable AI. Organizations must notify the public before implementing artificial intelligence systems, using accessible formats for all communications. People should be able to sign up for updates and provide feedback through accessible channels.
4. Data Quality and Bias Prevention: Before using data about people with disabilities in artificial intelligence systems, organizations must verify that the data is fit for purpose and includes appropriate representation of people with disabilities. This process should involve people with disabilities in decision-making and actively work to prevent bias, misrepresentation and unfair categorization.
5. AI System Design with Accessibility Built In: Accessibility and equity must be considered from the earliest stages of artificial intelligence design. Organizations must:
   • Test AI systems with people with disabilities throughout development
   • Actively solicit feedback from users with disabilities
   • Avoid using AI for surveillance, manipulation or categorization of people with disabilities
   • Establish clear accountability for AI decisions within the organization
6. Procurement of AI Systems: When purchasing artificial intelligence solutions, organizations must include accessibility and equity as key criteria. This requires testing systems before implementation, engaging disability experts to evaluate AI tools for equity and accessibility and including contract termination clauses if systems fail to meet standards.
7. Ongoing Monitoring and Harm Mitigation: Accessible and equitable artificial intelligence isn't a one-time implementation. It requires continuous monitoring. Organizations must regularly assess how AI affects people with disabilities, maintain a public record of harms and collaborate with disability community groups to establish risk thresholds.
8. Staff Training and Capability Building: Everyone involved in developing, deploying or managing artificial intelligence systems must receive accessible, role-specific training on accessible and equitable AI practices. Training should cover data privacy, universal design principles for AI, risk and harm identification, equity considerations and disability inclusion. People with disabilities should participate in designing and delivering this training.
9. Transparency, Consent and Data Protection: Organizations must provide clear, accessible explanations of how artificial intelligence systems and user data are used. They must share clear, accessible information about AI use and obtain explicit permission before using user data. Consent processes must be genuinely voluntary and people must be able to withdraw consent without penalty. Organizations must include people with disabilities in decisions about unconsented data collection. Throughout the data lifecycle, organizations must protect disability-related information and comply with privacy regulations, including the Privacy Act and PIPEDA.
10. Providing Human Alternatives: This is a cornerstone requirement. People with disabilities must always have access to human decision-makers or human review options that are equally available, timely and convenient. If an organization deploys artificial intelligence for critical decisions, it must maintain alternative pathways. For instance, deaf users seeking interpretation services must be able to request human interpreters and blind students taking exams must be able to request human proctors.
11. Handling Feedback and Complaints: Organizations must establish clear, accessible complaint mechanisms and respond promptly to feedback about AI accessibility and equity. They must accept anonymous feedback and explain what corrective actions will be taken to address reported issues. Organizations must keep people updated on the status of their complaints and offer transparent ways to review and challenge decisions made in response to feedback. When people with disabilities report harms caused by artificial intelligence systems, these reports should be added to a public record of harms (with appropriate anonymization and consent).
12. Continuous Improvement: Organizations must regularly review how artificial intelligence systems function and where they may be causing harm to people with disabilities. These reviews must include people with disabilities to ensure that real-world impacts are identified and understood. When AI systems fail to meet standards or cause significant harm, organizations must improve them or discontinue their use. Importantly, AI systems must learn from mistakes, including complaints of harm reported by people with disabilities. This iterative process ensures artificial intelligence systems evolve to better serve all users.

4. Accessible Education and AI Literacy

For artificial intelligence to be truly inclusive, everyone, including people with disabilities, must understand how AI works, how it affects them and how to advocate for accessibility and equity. Whether they build it, buy it, use it or are affected by it. Education and training must be accessible, teach about accessible and equitable AI, include people with disabilities in creating and delivering training and help people understand how AI affects their choices and independence.

1. Making AI Education Accessible: Organizations must use accessible tools and teaching methods when educating people about artificial intelligence. Training should help people with disabilities understand how to participate in AI governance and management and should support disability community members in all AI-related roles.
2. Training AI Professionals: Technical teams, managers and decision-makers must receive training on accessible and equitable artificial intelligence practices. This training should cover the requirements outlined in CAN-ASC-6.2:2025 and emphasize practical implementation of accessible AI and equitable AI principles.
3. Meaningful Disability Community Involvement: People with disabilities must help design and deliver AI education and training programs. This ensures that training reflects real-world challenges and incorporates disability perspectives from the start.
4. Helping Users Understand AI: Organizations must help people understand how artificial intelligence systems work and what these systems mean for their lives and autonomy. This includes explaining goals, benefits, risks and equity implications, helping people understand their choices regarding AI use, and teaching people how to give informed consent, select alternatives and challenge AI decisions.

Why CAN-ASC-6.2:2025 Matters for Your Organization?

Whether you work in technology, healthcare, finance, education, human resources or any other sector, artificial intelligence systems are likely affecting how you serve people with disabilities. CAN-ASC-6.2:2025 provides a clear roadmap for compliance with accessibility standards and equitable AI requirements. The standard helps organizations:

• Reduce legal risk by meeting accessibility compliance requirements under the Accessible Canada Act and Canadian Human Rights Act
• Improve AI quality by identifying and eliminating bias, discrimination and accessibility barriers in AI systems
• Build trust with users, employees and customers by demonstrating commitment to inclusive artificial intelligence
• Future-proof AI investments by ensuring that AI systems remain accessible as technology evolves
• Empower people with disabilities by centering their needs and preferences in AI design and deployment

How to Get Started with CAN-ASC-6.2:2025 Compliance?

Implementing the requirements of CAN-ASC-6.2:2025 requires organizational commitment and structured planning. Here are key steps:

• Assess Current AI Systems: Conduct an accessibility audit of all current artificial intelligence systems and tools your organization uses. Identify gaps in accessibility and equity features and prioritize improvements based on impact and risk.
• Establish Governance: Create or strengthen governance structures for accessible and equitable artificial intelligence. Include people with disabilities in decision-making roles and establish clear accountability for AI accessibility and equity.
• Develop AI Procurement Standards: If your organization buys artificial intelligence solutions, establish procurement criteria that require accessibility and equitable AI features. Conduct testing before implementation and include contract clauses for accountability.
• Implement Accessible AI Design Practices: If your organization develops artificial intelligence systems, integrate accessibility and equity considerations into every phase of development. Involve people with disabilities in design, testing and iteration.
• Build Staff Capability: Provide accessible training to all staff involved in AI work. Ensure people with disabilities participate in creating and delivering this training.
• Monitor and Iterate: Establish ongoing monitoring processes to assess how artificial intelligence affects people with disabilities. Create feedback mechanisms, maintain records of harms and use this information to continuously improve AI systems

Getting CAN-ASC-6.2:2025 Compliance Right with Expert Support

Implementing CAN-ASC-6.2:2025 is complex, but you don't have to do it alone. BarrierBreak combines AI-powered automation with human accessibility expertise to help organizations audit, manage and demonstrate compliance with the standard. From identifying accessibility gaps in your current AI systems with A11yInspect and accessibility testing, to continuously monitoring compliance with the A11yNow platform, to building team capability through accessibility training, BarrierBreak covers every phase of your compliance journey. With 20+ Years in Digital Accessibility, 300+ accessibility experts and 200K+ yearly testing hours, BarrierBreak has helped organizations across Canada and globally ensure their AI systems are truly accessible and equitable.