Consultation on computing power for artificial intelligence (AI) (Option consommateurs comments)

These include standards for the retention, destruction and anonymization of personal data, as well as rules governing consumer consent to the use of their personal information. In addition to the necessary infrastructure, artificial intelligence training requires a significant amount of data. This training data can come from consumers. It can be extracted from their online activities, such as profiles, posts and photos.

This data can sometimes be sensitive. Think of discussions about a mental health issue on an online forum, or a photo of a young child shared with friends and family. With the increasing development of artificial intelligence systems, the appetite for online consumer data is likely to grow.

This data may remain on the servers of the companies collecting it for AI training purposes, even after consumers have deleted it from the place where they originally posted it. Artificial intelligence systems may also disclose personal information used for training purposes. For example, researchers have been able to remove personally identifiable information, such as names, addresses and photos, from generative AIs. 

At present, companies wishing to train artificial intelligence but lacking the necessary infrastructure can already lease computing power from external service providers. These providers may be located outside Canada, where privacy standards may be less protective.

What's more, it can be difficult to identify the privacy practices of these companies through their terms of use or privacy policy. Indeed, these can sometimes be vague about the exact protection or use reserved for consumers' personal information.

In light of the issues outlined above, a Canadian-owned and -controlled artificial intelligence (AI) computing power infrastructure would better protect the privacy of Canadians.