(Version française disponible ici)

Which level of government in Canada has jurisdiction over artificial intelligence?

In the context of the division of legislative powers in the Canadian Constitution, asking this question amounts to asking who is responsible for managing modernity.

As far as artificial intelligence is concerned, several premises need to be established. First, there is a clear risk that digital technology will perpetuate or even exacerbate discriminatory practices, particularly through poorly calibrated algorithms. Current federal and provincial privacy laws are inadequate to meet the challenges posed by this new technological capability. Finally, while we must recognise that digital technology is now part of the economy, we must also acknowledge that economic growth will have to be accompanied by sound management of the use of technology.

A historical tug-of-war

The division of powers is at the heart of the dynamics of federalism. In 1867, the founding fathers wanted the provinces to have legislative powers over local and private issues. On the other hand, they entrusted the Canadian Parliament with power over the interests of the country as a whole or other matters that transcended the specific issues that belonged to the provinces.

In a 1937 ruling on a dispute between the federal government and Ontario, the Judicial Committee of the Privy Council found that the division of legislative powers were exclusive prerogatives. “While the ship of state now sails on larger ventures and into foreign waters she still retains the watertight compartments which are an essential part of her original structure,” wrote Lord Atkin.

More recently, the Supreme Court of Canada has put the principle of “watertight compartments” into perspective by authorizing the simultaneous presence of both levels of government in the same constitutional space. As stated in a 2015 judgment in a case between the federal government and Quebec, the “modern view of federalism favours a flexible conception of the division of powers. . . allowing governments at both levels to legislate for valid purposes in the areas of overlap.”

This co-operative federalism should apply to artificial intelligence, which lends itself particularly well to shared jurisdiction, following the example of health care or the environment. On the federal side, it is mainly powers over trade and commerce and criminal law that would apply. In the case of the provinces, it could be powers relating to property and civil rights, or to matters of local or private concern.

Resisting the one-size-fits-all temptation

Some would like to apply the residual jurisdiction provision, whereby the Constitution entrusts by default to the federal Parliament everything that does not immediately fall within the powers of the provinces. Artificial intelligence is not a fit for residual jurisdiction: AI is not a new matter, but rather an aggregate of both provincial and federal matters. We also note that most of the residual federal powers support or extend powers already assigned to the federal government. These powers are therefore less and less  distinct areas of federal jurisdiction, and more of an explicit use of existing jurisdiction.

Other observers might invoke the Canadian Parliament’s jurisdiction over issues of national scope or interest. We do not believe that this argument applies, since artificial intelligence does not possess the characteristics of “singleness, distinctiveness and indivisibility” that are required according to the Supreme Court of Canada. Moreover, it is difficult to see how the federal government could apply the principle of provincial incapacity, which the Supreme Court requires to be demonstrated in the application of the national interest theory.

In reality, artificial intelligence is a matter of dual aspect, meaning both the provinces and the federal government could legitimately seek to regulate it, at least to a certain point.

As mentioned above, there is probably no subject more suited to the application of “cooperative federalism” than artificial intelligence. The federal and provincial governments have no choice but to work together to develop pan-Canadian standards and common requirements. It is only natural that they should join forces in the international competition currently underway to define the rules for artificial intelligence.

Is Bill C-27 constitutional?

Last year, the federal government tabled Bill C-27, which aims to regulate artificial intelligence and data. Does it respect the spirit of co-operative federalism?

As expected, the bill is based primarily on federal powers in the areas of trade and commerce, protection of personal information and criminal law. It has two main objectives. First, to regulate trade and commerce in artificial intelligence systems and the data generated by these systems and, second, to prohibit conduct that may cause serious harm to individuals or prejudice their interests.

Ottawa’s bill, thus circumscribed, constitutes a valid exercise of the federal powers enumerated in subsections 91(2) and 91(27) of the Constitution Act, 1867.

As far as the federal spending power is concerned, it is not really at issue, since we are talking about the adoption of rules, not expenditures. However, Ottawa could one day use its spending power to support its interventions in the field of artificial intelligence. The Supreme Court considers that this power can be exercised independently of the division of legislative powers.

No level of government can or should claim total and exclusive jurisdiction over the here and now. Artificial intelligence is not the fruit of spontaneous generation, but rather the result of technological evolution. It would be completely opposite to the spirit of the Canadian Constitution for any technological advance to fall automatically under federal jurisdiction. As far as the present world is concerned, the Canadian provinces also have a say.

Despite appearances, C-27 is very timid when it comes to artificial intelligence, which would leave Quebec and the other Canadian provinces all the space they need to assume real leadership in this area.

Read more in this series:

• The ethics of artificial intelligence await the law, by Jocelyn Maclure and Alexis Morin-Martel
• The time for a law on artificial intelligence has come, by Céline Castets-Renard and Anne-Sophie Hulin
• The risk of waiting to regulate AI is greater than the risk of acting too quickly, by Jennifer Quaid
• How to legislate on artificial intelligence in Canada, by Céline Castets-Renard and Anne-Sophie Hulin

THE IRPP IS HOLDING A (FREE) WEBINAR ON ARTIFICIAL INTELLIGENCE ON OCTOBER 5 AT 1 P.M. ET

Click here to register!