Abstract:
As AI becomes increasingly embedded in Canadian systems and society, this panel explores how to govern it responsibly post-deployment. Experts in law, policy, ethics, industry, and Indigenous sovereignty will discuss policy and regulatory tools to manage AI risks, promote equity, and build public trust. The dynamic session includes Think-Pair-Share, live polling, and a word cloud to engage attendees in shaping actionable solutions. It concludes with a call to action and a synthesis of audience-driven priorities to guide Canada’s leadership in safe, accountable AI.

Summary of Conversations

Discussion focused on the risks and harms that can arise from the deployment of artificial intelligence (AI), prompted by three use scenarios:  employment screening, Indigenous language preservation, and election deepfakes. A central theme arising in the discussions was the pervasive risk of bias and misrepresentation. Participants underscored that AI reflects and entrenches existing systemic harms and societal biases present in training data. For culturally sensitive applications, such as Indigenous languages, critical concerns were raised regarding data sovereignty, the necessity for collective consent, and the potential for cultural distortion. The conversation repeatedly highlighted a widespread erosion of public trust due to a lack of transparency and the difficulty for citizens to discern authentic from synthetic content. A strong consensus emerged that current legal and regulatory mechanisms are wholly inadequate and under-resourced to address these complex, evolving harms.

Take Away Messages/Current Status of Challenges

• Amplification of Systemic Bias: AI systems, trained on historically flawed human data (e.g., past hiring decisions), can perpetuate and even escalate existing societal biases and systemic harms at speed and scale.
• Legal and Regulatory Obsolescence: The existing legal and regulatory frameworks are not equipped to govern AI, featuring outdated privacy laws and an absence of a comprehensive, modern legal structure for recourse.
• Erosion of Public Trust: The deployment of AI is characterized by a lack of transparency and disclosure, which fundamentally erodes public trust in institutions, media, and technology.
• Under-Resourced Enforcement: Public institutions tasked with upholding anti-discrimination and privacy rights (e.g., human rights commissions) are chronically underfunded and lack the necessary resources and expertise to handle technologically advanced AI complaints.
• Complex Data Sovereignty Issues: For low-resource and culturally situated data, such as Indigenous languages, there are major barriers related to securing collective and informed community consent and ensuring data sovereignty.
• Shifted Burden of Discernment: The public bears an unfair burden to acquire specialized AI literacy to detect advanced synthetic media (deepfakes) and misinformation, equating trust with an individual’s technical ability to discern authenticity.
• Insufficient Policy-Making Inclusion: The policy development process for AI governance has largely excluded critical voices from the public, consumers, and academics, resulting in proposed legislation that is criticized as inadequate and industry-centric.
• Opacity Hinders Accountability: The lack of clarity around AI operations means that deployers, users, and affected individuals often “don’t know what’s going on,” which makes establishing accountability exceptionally difficult.

Recommendations/Next Steps

• Mandate Proactive Transparency: Implement policies that enforce strong disclosure requirements to ensure full visibility and understanding of AI deployment and its functional mechanics for all stakeholders.
• Introduce Explicit Legislative Duties: Establish new, positive legal duties on AI developers and deployers, including a duty of care and a prohibition against unjust discrimination, leveraging well-established legal concepts. The focus should be on harm reduction as well as risk reduction. Legislative duties should also address mechanisms for the public to flag and voice harms.
• Prioritize Public and Affected Community Consultation: Require far more comprehensive and genuine public participation and pre-deployment investigation to anticipate, appreciate, and mitigate potential harms before systems are implemented. The public should be engaged as citizens and not merely as consumers of AI technology. The race to AI has been commercially driven at the expense of human rights consideration.
• Create Dedicated Recourse and Complaints Channels: Urgently develop and fund new, effective mechanisms and specialized avenues where individuals can report problems, raise complaints, and receive timely responses regarding harms from AI systems.
• Build Sovereign Digital Infrastructure: Strategically invest in and develop domestic AI technology and data infrastructure to ensure that models reflect national ethical values and that data is processed under local laws. Canada should lean into AI development and deployment and not defer to foreign countries to define what regulation looks like for our citizens and our nation.
• Implement a Supply Chain Responsibility Model: Adopt a framework that enforces responsibility across the entire AI lifecycle, obligating model developers to test for bias and holding deployers accountable for the system’s eventual use.
• Rigorously Fund Enforcement and Expertise: Substantially increase the funding, staffing, and specialized technical expertise of human rights and privacy enforcement institutions to enable them to effectively litigate and regulate complex AI-related issues.

Accelerate Modern Legislative Action: Expedite the political will and process to enact known, effective legislative standards for AI governance, drawing lessons from international frameworks to ensure a timely legal foundation.