Transcript
Speaker: Hello and welcome to today's episode of Red Eye. I'm Jane Williams. In this episode, Tim McSorley joins Lorraine Chisholm to talk about why he and other human rights activists are urging MPs to vote against the federal government's proposed legislation on artificial intelligence.
Speaker: Every day we hear about new and widespread applications of artificial intelligence, from facial recognition to deepfakes to chatbots. In Canada, leading privacy technology and civil liberties experts are urging Canadian Members of Parliament to vote against the federal government's proposed Artificial Intelligence and Data Act, which would regulate AI. Signatories to an open letter to Members of Parliament say it is flawed beyond repair.
Speaker: Tim McSorley is National Coordinator of the International Civil Liberties Monitoring Group, and he joins me this morning. Hi, Tim. Wonderful to talk with you again. Great to talk with you too, Lauren. Thanks for having me on. To begin with, though the letter to opposition leaders calls for the defeat of the proposed Artificial Intelligence and Data Act, you're actually calling for very strong regulations on AI. Why is this type of regulation necessary?
Speaker: Well, right now what we see in Canada is that artificial intelligence tools that pose a really grave risk and high risk to privacy, the human rights and civil liberties of people, not just in Canada, but anyone who might
Speaker: might have access to or might have to be involved with those AI tools is completely unregulated. So right now we know that in the private sector, for example, that anyone from banking and healthcare to online analytics companies are all using AI to either evaluate risk
Speaker: to develop profiles, to simply use AI tools, for example, that would chatbots that make their work easier. But as these grow, they're used in more and more sensitive ways and more universal ways.
Speaker: And so it really has, is starting to have an impact on our day -to -day life. And just, you know, in the months since this bill was introduced and just even a little over a week since we sent this letter, I think everyone's probably seen how chat GPT and other AI tools are really, you know, taking us by storm and being used in all kinds of ways that even just a month ago, I don't think people who weren't following this and even some of us who were really knew would be happening.
Speaker: And so the fact that we don't have strong regulation is an issue. And our concern is that the way the government has gone about introducing these regulations in ADA means that they won't be able to get the full study that it deserves because it's part of a much larger bill.
Speaker: And so we think that the government needs to start over. And the reason we're asking them to leaders and MPs to vote against this part of the overall bill is so that we can have a dedicated piece of legislation that reflects better the concerns that many civil liberties and human rights and technology experts have already expressed introduced into parliament and that that formed the basis for regulation going forward.
Speaker: Well, let's walk through your critiques of the bill as it exists, starting with who was consulted. What happened on that front and who was left out? So while there was a sense that there were regulations coming and that there have been groups that reached out to the government to share their concerns around what should be in regulations and the need for them.
Speaker: We as a coalition and those who signed on to the letter and that we've spoken to were not consulted. So individuals working in civil society, organizations, academics who are studying it. And from what we understand, even most of the major industry groups also weren't appropriately consulted. So that's one of the primary concerns is that the government, despite knowing that there are people and organizations who have
Speaker: being spending countless hours thinking about what these regulations could look like, that they didn't engage in a thoughtful, formal, and in -depth consultation process before coming forward with the bill. I also understand that many important pieces of the Act have yet to be detailed out. Why is that a problem?
Speaker: Right. So the goal of the bill is to regulate what they call high risk AI. And so even the idea of what constitutes high risk would be left to regulation definition by regulation after the bill is adopted.
Speaker: And so, you know, things like that, how to assess, you know, what criteria needs to go into assessing whether or not a piece of artificial intelligence is high risk would also be left to regulation and what kind of safeguards an organization would have to put in place if they do have and produce high risk AI tools. And so the fact that all of these things would be left to a regulation process after the bill is adopted,
Speaker: is very concerning. The government's argument is that the process for forming regulations is public and it's true and they could have consultations after the bill is adopted that they didn't have before introducing this legislation in order to get into what the regulations would look like.
Speaker: But at the same time, our experience is that the regulation process is much more obscure, involves much less consultation, and isn't as visible to the public in the way that it would be when it's going through Parliament and being debated by MPs. And so we don't think it's adequate to say that we'll define this by regulation after the fact.
Speaker: Now the groups calling for defeat of ADA are concerned about regulation and enforcement in particular, if the bill were to pass. Why is that so important? So the bill proposes to create a new position called the artificial intelligence and data commissioner, and that they would assist the minister in administering and enforcing the act and the regulations, which is a good thing. We agree that there needs to be a dedicated office to do this.
Speaker: The concern, though, is that the individual is appointed by the minister and isn't an officer of parliament. So, for example, the Privacy Commissioner of Canada, the Information Commissioner and others, they aren't answerable to a particular minister. They don't fall under that minister's office. And our concern would be that the minister of industry would be able, because it's a political position,
Speaker: that there wouldn't be that arm's length nature in how it's reviewed and how it's enforced, that there could be concerns around political interference or patronage appointments, and simply that if we want to have clear appearance of a separation between the office that would be enforcing this and the ministry that's developing the rules and regulations, that they need to be more independent.
Speaker: And so one option would be to create a whole new position called the Artificial Intelligence and Data Commissioner. But there's also being raised the question of whether or not something like this could fall under the powers of the Privacy Commissioner of Canada, because it is so related to privacy concerns. Although at the same time, we know that the Privacy Commissioner already has a lot of issues to contend with, a lot of work to do, and itself is looking for more resources.
Speaker: creating a new position wouldn't be a bad idea, but it's just the way that they're going about it, we think is inadequate. Now, one thing that seems particularly concerning is that the act doesn't apply to government institutions like the RCMP. Why are you arguing that it should?
Speaker: Just to be clear, the point of ADA is to regulate the private sector. So that's one concern is that we have this private sector regulation coming in, and we've even promised that there will be public sectors, so government and crown corporations, that regulations for them will be coming later. We assume as part of a package to update the Privacy Act, which is what governs the privacy laws in the federal public sector.
Speaker: So in the meantime, this doesn't apply to government agencies, which we think is too slow. And we think that there could easily be one bill that would lay out the regulations. And we acknowledge that there could be differences in how AI in the private sector and AI in the public sector and within the federal government are regulated slightly differently. But that could be done all at the same time. But then even given all that, there's an exclusion in this act that says that
Speaker: AI tools that are directly under the control of national security agencies are excluded. And so what that means and what we've been told it means is that if a company is exclusively making an AI tool for the RCMP, for the Canadian Security Intelligence Service and for others, that these rules wouldn't apply to them. And we think that's incredibly troubling.
Speaker: On one hand, it signals a greater concern that we have and that we've seen in other jurisdictions that national security agencies in general wouldn't be covered by either private sector or public sector AI regulation. So for example, in Europe right now, under EU regulations, even for the public sector, they're excluding national security agencies from being covered by it. And we know Canada is modeling their regulations off of what the EU has been looking at.
Speaker: And then we also simply don't believe that tools created by the private sector for government use will stay with the government and that it won't somehow elude proper regulation when it does eventually go out to the private sector, because that's what we see that tools developed for government institutions don't just stay there that, you know, these private companies find a way to monetize it in other ways.
Speaker: And that eventually it would get out into the wild, you know, so to say. And we also just think that, you know, these private companies are still creating AI tools that they're making a profit off of that they're selling to the government. They're still in the private sector and that they should be covered by the same rules and regulations that apply to other private sector development of artificial intelligence tools.
Speaker: Many of us have been very alarmed by how facial recognition has been used. And, you know, and there we are talking about police institutions in Canada. And you point out that the act does not address the significant human rights implications of algorithmic systems. It's a huge topic in itself. We could do a series on it. But could you sketch out for us some of the real dangers here? Certainly, yes. What we've seen with facial recognition technology is that
Speaker: Both the algorithms that most facial recognition technology are based off are biased against people of color and racialized people and women. And so there's a high level of inaccuracy. I have to say that we know that the companies that create these tools have been working very hard to address the inaccuracy issue because in the end they realize that that's not very good for their business if what they're marketing is something that's supposed to be accurate.
Speaker: But even if these tools are more accurate than not, there's the concern that AI systems and AI tools such as surveillance would continue to be used in problematic ways as we see with surveillance in general against racialized communities in particular. We only have to think about the over surveillance of black neighborhoods and black communities across Canada or the over surveillance of Muslim communities, you know, by Canada's anti -terrorism organizations.
Speaker: And then it's also used in other ways. So for example, for facial recognition, we know it's being used to try to identify and match photos of asylum claimants when immigration Canada doesn't believe they are who they say they are.
Speaker: and using algorithm and facial recognition tools to try to match the person's face to other photos they may have on file, other publicly available images. And that's been highly problematic. And so there's a high impact on people's human rights and on their sensitive information. And not once in ADA does a talk about taking a human rights approach of ensuring that the assessment of what is considered a risk is based in a human rights assessment.
Speaker: or that groups that work on human rights and civil liberties concerns or that work within the communities that are most affected would be involved, for example, in the consultation process around assessing whether or not a tool is high risk. So we think that this is a clear black in the act and something that could have been there from the beginning.
Speaker: that the government, despite not having consultations, has been informed of these concerns by people in the past, and they simply chose not to include it. What can folks do to make sure that it is voted down in its present form and we get regulations for artificial intelligence that address these very important concerns?
Speaker: Right now, I think the best way they can express their concern would be to reach out to their local MPs. It's currently being debated in the House. The timing is good. Starting Friday, the House is on break for two weeks. So MPs are supposed to be generally going to be back in their ridings to meet with constituents. So to reach out to your Members of Parliament to raise concerns. And if you want, you can go on our website to share the letter with them as well. One thing I'll say is that
Speaker: Based on where things currently stand, we are asking MPs to vote against this bill at second reading. But one thing that's come out since we sent the letter is the possibility and the importance of putting pressure that if the government were to formally split the two. So the situation we're in right now is a very kind of technical parliamentary situation where there's going to be two votes, but the bill isn't split. So the only option would be to vote against the section about ADA.
Speaker: and then it would just disappear, and the government would have to reintroduce other legislation. But they do have another option that we've been made aware of, and that's under consideration, and that's to formally split the bill so that ADA would be its own standalone bill, and then there would be more of a possibility of bringing in changes if it had its own study at committee. So the message that I think is important to send right now is that if things don't change, the parties need to vote down, ADA, and we need to start over from scratch.
Speaker: But if the government is willing to take a second look at this, to really split the bill and have standalone study of their legislation on artificial intelligence, that maybe we'd be able to fix it. You know, we'd still have to see what happens there and then there's always a chance it still isn't adequate. And then there's a possibility to vote against that at third reading. But really our goal is to find a way to have strong and enforceable AI regulation.
Speaker: And it really does have to happen quickly because we just constantly see new ways and new developments in artificial intelligence where it's becoming much more pervasive in our society in good ways and bad. And we need to make sure that there are rules and legally enforceable rules around how it's developed and how it's used.
Speaker: Well, thanks so much for joining me, Tim, and I'll point people to where they can find the current open letter and encourage people to write to their MPs. Great. Thanks so much, Lorraine. I've been speaking with Tim McSorley. He is National Coordinator of the International Civil Liberties Monitoring Group. The open letter can be found at their website, which uses their initials. So ICLMG .ca. I'm Lorraine Chisholm.
Speaker: The Red Eye Collective is based in Vancouver. You can check us out at coopradio .org slash redeye.
