There was a lot of concern coming into the last election that we would see—and not just in Canada but also in other elections around the world—deepfakes as a major threat. We have seen deepfakes, but I don't think the threats or the dangers have yet materialized.
I think you've heard witnesses before me who are of the view that it continues to be a risk. I share that view The manipulation of voices and images could deeply mislead electors, and the sophistication that comes with AI increases that risk. It's very cheap. It's easily accessible to anybody. We have to expect that it's going to come.
I don't know that there's a magical solution to deal with that, but there are elements. I do believe that it would be appropriate to have a watermark, basically a marker, whenever AI is used in the production of electoral communications. At the very least, if it becomes unlawful to not put that label, platforms' own moderation policies, which typically include taking down unlawful content, could act upon that immediately. It creates additional leverage. Of course, it does create more transparency for Canadians.
I also think that there should be some very narrow but specific prohibitions on misrepresentations of political actors or of the Chief Electoral Officer, with regard to using or manipulating by any means the voice or the image to have them say things they haven't said or put them in a context where they haven't been.