Industry Committee on Oct. 17th, 2023

[Mark Schaan]

The privacy link between Bill C‑27 and the EU's regulation is important in a few ways. The bill lays out some important definitions and elements to support interoperability between the two pieces of legislation.

As far as alignment with the EU is concerned, we are engaged in an ongoing process with our EU counterparts. It began three years ago, and we are confident it will result in the right connections between the two pieces of legislation.

[The Chair, Liberal, Joël Lightbound]

Thank you.

Go ahead, Mr. Masse.

[Brian Masse, NDP, Windsor West, ON]

Thank you, Mr. Chair.

Thank you for being here again.

On September 27, the day after the minister was here, he announced a voluntary code of conduct on the responsible development and management of advanced generative AI systems. Can you tell us more about that initiative and who's involved and who isn't?

I was part of creating the right to repair bill. It's a voluntary agreement. Some didn't even join the agreement and there are issues with it.

If you could give us a bit of background on that, it would be appreciated.

[Mark Schaan]

Thank you, Mr. Chair, for the question.

The operating context for the code on generative AI is important, because it actually falls within a broader milieu that I think is useful. At the digital ministerial of the G7 in Gunma, Japan, earlier this spring, digital ministers of the G7 recommended to leaders that they undertake work related to generative AI and that they come to some conclusions that could be shared with ministers.

Then, at the Hiroshima meeting of G7 leaders, this was formalized in a call from G7 leaders to their respective digital ministers to do work related to generative AI, notably work on aspects of principles for governance of generative AI systems and potential work on a code of conduct. That prompted a lot of work between Canada, which was very active in the development of the actual recommendation from ministers, and our international colleagues around some of the work that needed to be done around guardrails for generative AI.

An early adopter, or an early mover in some ways, was the United States. Our counterparts, under the Biden-Harris administration, produced a set of commitments from their largest foundation model developers to a certain set of standards that they would hold themselves to in the development of foundation models, or what's sometimes called generative AI.

Canada recognized that we not only wanted to be a participant in the Hiroshima G7 process and have value added in terms of the discussion around principles and the code of conduct, but we also have our own domestic recognition of what would be important guardrails. That's particularly because AIDA, the artificial intelligence and data act, has regulations that will take shape in many ways, but the potential for concerns and harms related to generative AI and foundation models is already present in many instances.

That spawned a set of discussions with the AI ecosystem, particularly those that were actively involved in the deployment of foundation models and generative AI within their private sector and commercial context. We held a number of round tables ultimately leading up to the development of the voluntary code, which was announced at the All In conference in Montreal. It depends on how one counts, but the signatories now include a number of key foundation model companies in Canada, notably Cohere, Coveo and Aida. It also includes endorsement from the Council of Canadian Innovators, which represents a much larger swath of leaders, as well as a number of some of the larger ecosystem players like OpenText and BlackBerry.

[Brian Masse, NDP, Windsor West, ON]

Are there any holes or are we missing certain sectors that need to be brought in on this? I haven't looked; I'll be looking in the future. I haven't had time. Are we missing a certain component of the sector, which we need to shore up?

Second, the problem we have with some of the bill here, as I understand it, is that some stuff will be left to regulatory powers. It sounds like this actually crosses over to some of that, potentially. How much of the code of conduct is actually going to be in the regulations of the bill? Also, going back, are we missing any holes with regard to building this voluntary arrangement?

[Mark Schaan]

As it relates to the development of the voluntary code, a key component was obviously that those who would actually implement it and be signatories to the implementation of the code were principally involved in its structure.

That said, we also went broadly to a number of academics and civil society actors. I think we're still very open to the conversations around its function, particularly because the G7 Hiroshima process continues. Canada will continue to work on both the principles to the development of AI foundation models and the code of conduct that we imagine will emerge internationally through the G7 process.

As it relates to the relationship between the code and regulation, we have been clear that the code is not regulation and the code will not be regulation. This is a bridge between now and when regulation will be in place.

That said, it would be silly of us—and it was not our intent—to make the code anathema to where we would ultimately head. Its principles are actually related to the principles and the general pillars that are related to the companion document we produced on AIDA. The general headings are the same.

In many ways, therefore, signing on now and and beginning the important work of actually doing the risk assessment, mitigation, transparency and accountability work related to generative AI will put companies in very good stead when we actually have regulatory systems in place.

[Brian Masse, NDP, Windsor West, ON]

We were able to get some agreements internationally on human cloning and things of that nature.

Is this kind of like the precursor to trying to get some international understanding on the deployment of AI that goes beyond? Is this where we're trying to head with this?

[Mark Schaan]

I think it's a recognition of the fact that frontier technologies are often global by nature, in that the use of data, the transcendence of data and the continued function of models in a borderless digital economy require a certain commitment to interoperability as well as a certain calling out of those who are not necessarily playing within that realm.

I think the goal of the G7 AI process is very much to ensure that we have effective international co-operation on these guiding principles and foundational elements, recognizing that it will look different in domestic contexts in terms of how they actually import that to either regulation making or legal precedent. In the EU case, that's the EU AI Act, as well as other functions of their own legislation. In Canada, that would be AIDA. In the United States, that's the Biden-Harris commitments, as well as what we understand through the press to be the forthcoming executive order on artificial intelligence.

[Brian Masse, NDP, Windsor West, ON]

Thank you.

Thank you, Mr. Chair.

[The Chair, Liberal, Joël Lightbound]

Thank you very much, Mr. Masse.

We have about 10 minutes, so I'll go three, three, one and one.

Mr. Williams, go ahead.

[Ryan Williams, Conservative, Bay of Quinte, ON]

Thank you.

Quickly, why was privacy not included as a fundamental right in the first draft of this bill?

[Mark Schaan]

Mr. Chair, I thank the member for the question.

I think we have continued to hear how best to reflect the nature of Canadians' rights in the bill itself. I think the important feedback of the Privacy Commissioner was fundamental in actually calling for the insertion of a fundamental right.

There have often been tricky questions about how the federal government engages in questions that actually transcend some of its roles in the economy. This is a private sector commercial bill and, fundamentally, I think we were very encouraged by the feedback from the Privacy Commissioner to contemplate its actual insertion.

[Ryan Williams, Conservative, Bay of Quinte, ON]

You met with the Privacy Commissioner prior to drafting this bill. Is that correct?

[Mark Schaan]

We have an ongoing engagement between the Office of the Privacy Commissioner and both my secretary and my branch because we have ongoing interests in a number of pieces of legislation.

[Ryan Williams, Conservative, Bay of Quinte, ON]

Did you meet specifically about this bill prior to the draft?

[Mark Schaan]

I have met with the Office of the Privacy Commissioner with respect to Bill C-27.

[Ryan Williams, Conservative, Bay of Quinte, ON]

We'll have him here on Thursday, but did he recommend specifically before—he's made comments since, as the minister has—that privacy should be included as a fundamental right in the bill?

[Mark Schaan]

The Privacy Commissioner issues an annual report and also publishes a letter to this committee with respect to his findings on the bill. I think you'll see the—

[Ryan Williams, Conservative, Bay of Quinte, ON]

I'm just asking for a yes or no. Did he give recommendations prior to the bill being drafted, when you met with him?

[Mark Schaan]

Prior to the bill being drafted—

[Ryan Williams, Conservative, Bay of Quinte, ON]

I mean, through consultations when you were putting this bill together.

[Mark Schaan]

We had engagement with the Privacy Commissioner in the lead-up to the tabling of Bill C-27.

[Ryan Williams, Conservative, Bay of Quinte, ON]

Whose definition of “minor” will be used when enforcing this bill?

[Mark Schaan]

Right now, as you'll note, there is no definition of “minor” in the bill. I imagine that's a question this committee will continue to contemplate as it studies the bill.

[Ryan Williams, Conservative, Bay of Quinte, ON]

I see the issue that it potentially creates 13 different enforcements of the protection of children's privacy, as the definition of a minor changes from province to province.

Was there any other reason why this was left undefined?