Thank you, Chair and members of the committee, for this invitation to appear today. I was recently asked to contribute to the federal AI strategy task force on the theme of trust and safety. My remarks today draw on that submission, which is publicly available.
I understand that this study is focused on AI research, on recent advances, on the needs of research institutions and on the role of the federal government in promoting a responsible AI research ecosystem. I want to address that final point directly because I personally think it's foundational to all the others.
The government has made it clear that it wants Canadians to adopt AI. This is understandable. I increasingly believe that this technology is transformational and has the genuine potential to improve productivity, reshape public services and strengthen our economic competitiveness. To do this, Canada should certainly invest in AI research infrastructure, fund centres of excellence and recruit world-class talent, but investment alone is not a strategy for adoption. Creating the conditions under which those systems can be safely integrated into public life is equally essential. If the systems that emerge from the Canadian AI research and development are deployed without clear standards for safety, transparency and accountability, and if the public does not trust them as a result, then the entire enterprise is undermined. A responsible AI research ecosystem, then, requires a responsible AI governance ecosystem. The two can't be separated.
Right now, that public confidence is simply not there. Only 34% of Canadians are willing to trust AI systems. Nearly 80% are concerned about negative outcomes on their lives, and 78% fear AI will undermine our elections. Seventy per cent are worried about their children's safety, and close to 80% believe AI threatens their jobs. However, 88% want stronger governance. This is not, then, a literacy gap. Canadians are making reasonable judgments. These systems were deployed without meaningful oversight, and they have seen and felt the consequences. This is a governance gap, and without closing it, neither the adoption the government seeks nor public confidence in AI research will materialize.
My submission to the task force argues that closing this governance gap requires action on three fronts. I think of these as the democratic foundations of AI governance—really the bare minimum we need to do.
The first is citizen safety. AI systems are being deployed at an unprecedented pace, often without meaningful risk assessments. We've seen chatbots fail users in mental health crises and children exposed to harmful content through many AI-powered products. Democratic governance requires independent regulatory authority with the power to mandate risk assessments before deployment and ensure heightened protections for children.
The second is information integrity. AI now both curates and creates our information environment. Synthetic media is proliferating. Provenance is obscured, and citizens cannot distinguish meaningfully between authentic material and fabrication. Democratic governance requires mandatory AI labelling and provenance disclosure, at least for now, and data access frameworks that allow research to study these systems.
The third is democratic legitimacy. Citizens must have meaningful agency over AI systems that affect their lives, recourse mechanisms when AI causes harm, data portability so that users are not locked into products, and structured public consultations so that governance is shaped by the people it affects.
These principles can be translated—and I would suggest easily—into concrete policy. Canada does not need sweeping new AI legislation. Most of these measures can be implemented through two existing frameworks, an amended online harms act that brings AI chatbots into scope and an amended consumer privacy protection act with AI-specific transparency requirements. Both bills had cross-partisan support. Both could be revived quickly.
I would also add that addressing this governance gap is itself a research challenge. Too often, AI funding is understood narrowly as investment in computer science and technical infrastructure, but understanding AI's effect on society and developing policy tools to address that require sustained investment in social science, humanities and interdisciplinary research. Canada has strengths in this area, but it remains significantly underfunded. If this committee is considering the needs of Canada's AI research ecosystem, I would urge you to think broadly about what that ecosystem must include.
Governance is not a constraint on a responsible AI research ecosystem. It is a precondition for it.
Thank you. I look forward to questions.