Thank you for the invitation to address the committee today.
I'm Sam Andrey. I'm the managing director of the Dais, a think tank at Toronto Metropolitan University where we work to develop the policy ideas to advance an inclusive, innovative economy, education system and democracy for Canada.
I'm going to focus my remarks today on the AI and data act. As many of my colleagues have noted, AI has the potential to have a transformative impact on our economy and our daily lives, but it also poses significant risks, including systemic forms of discrimination, psychological harms and malicious use.
The latest data from StatsCan shows that only about 4% of Canadian businesses are using AI, so to reach AI's full potential and increase adoption, we need a responsible governance framework.
Unfortunately, we think the current bill fails to adequately do that. The bill's surprise introduction and lack of public consultation since have limited the ability of folks in civil society, experts, industry and equity-deserving communities to engage with this important legislation. Our team at TMU, led by Christelle Tessono, has partnered with McGill University's Centre for Media, Technology and Democracy to engage with many of these folks over the last year and has produced recommendations for improving the bill, which we'll be sending to the committee.
I'm just going to highlight three of those today that we hope can be addressed if AIDA is moving forward.
First, the bill's definition of “harm” is very narrowly focused on individuals, but the harms of AI systems also occur at broader community and group levels. Depending on the type and context of the system in question, harm to individuals can be difficult to prove and only evident when assessed at a population level. Moreover, there are types of collective harms that are manipulative and exploitative from AI that would likely not be captured by this definition. Things like election interference, harm to the environment and collective harms to children are not harms that would be captured by the definition, which is focused on individuals.
Second, as my colleagues have said, the proposed regulatory model does not create sufficient independence from the minister of ISED, who would have competing roles of championing the economic benefits of AI while regulating and enforcing its risks. We think that the proposed AI and data commissioner needs to be independent from the minister, ideally through a parliamentary appointment and certainly with sufficient resources to support their role.
We would also propose two additions. One is the ability for individuals to make complaints to the commissioner. Currently to launch any investigation, the minister has to have reasonable grounds to believe that an investigation is warranted, which is a very high bar. The other is for the commissioner to be able to conduct pre-emptive audits.
Third, as has been mentioned, this bill currently only applies to the private sector. Minister Champagne's proposed list of high impact systems that he's shared with this committee that would be potentially subject to regulation includes a number of AI systems commonly used by public sector actors, like facial recognition used by the police and health care, but it creates a double standard where the private sector developers of these systems are going to be subject to regulation and our public servants operating them will not be.
This double standard is unlike the EU, and it fails to position the Canadian government as leading by example through legal bans and guardrails for its own responsible development and use of AI. The current structure of the bill, particularly its commissioner being an ISED departmental official, makes it poorly structured to provide oversight for all public sector AI. We acknowledge that it would not be an easy amendment job, but I would just note that Parliament needs to prioritize the development of AI regulation for the public sector, which needs to include adequate public consultation and engagement.
I want to close by saying that Canada's investments in developing AI systems and research have not yet been matched by a comparable effort to regulate the quickly evolving risks of the technology. We're encouraged that the minister and this committee are open to amendments that will strengthen the bill, and there's really a large community across Canada who wants to help.
Thank you for the opportunity.