Policy Matters: 5 Measures to Fix Canada’s Proposed AI Legislation

The $2.4 billion package for artificial intelligence (AI) included in the 2024 federal budget is a positive step for Canada’s AI ecosystem — but maintaining our competitive advantage in the space requires more than dollars. It’s critical the federal AI regulatory landscape aligns with funding. If passed as is, Canada’s proposed AI legislation (Part 3 of Bill C-27) will unintentionally limit investment and adoption by businesses.

And at a time when Canada’s lagging productivity and declining global competitiveness have reached such extremes that the Bank of Canada has declared the situation a national emergency, Canadian businesses can’t afford to sideline adoption of artificial intelligence. 

While there’s no question that government has a clear role in safeguarding the public through thoughtful AI legislation, Bill C-27 should provide Canada a clear legislative model for regulating AI that balances its risks and benefits. Currently, it does not.

In March, the Canadian Chamber, on behalf of our members, submitted a letter to the House of Commons Standing Committee on Industry and Technology with recommendations for modifying Bill C-27. Here are five of our recommended fixes:

Provide a clear definition for “high-impact systems”

It is important that the bill includes a definition of “high-impact systems” instead of using a broad classification approach that relies on scenarios and use cases — some of which catch “low-impact” systems in their net.

A clear definition will provide businesses with certainty regarding their obligations.

The bill should define high-impact so that it is focused on consequential uses that have a material, legal or similarly significant harm to housing, employment, credit, education, healthcare, criminal justice or insurance.

Don’t regulate General Purpose AI (GPAI) systems based on our current understanding of capabilities

Since the risks of GPAI systems are context dependent, the mitigation strategies should also be context dependent.

Developers of GPAI systems can’t anticipate every possible application of their product and the associated risks. Downstream users are in a better position to comply with the obligations that arise. Bill C-27 should regulate GPAI based on its role in the supply chain.

It’s important our legislation recognizes that we don’t have the full picture of AI and all its possible applications yet. By designing technologically neutral and risk- and principles-based legislation, we can create sustainable regulations amidst a time of rapid change that will apply equally to emergent technologies.

What does “principles-based” mean?
The use of high-level, general statements/principles — rather
than specific rules — that apply across a broad range of
situations and focus on outcomes.

Ensure alignment with our international trading partners

Work on AI is happening at a global level. It’s in Canada’s best interest not to rush ahead with legislation that could put us out of step with major trading partners.

Bill C-27 should specify that forthcoming regulations must align with international AI-governance standards and promote interoperability (when a product/system can work with, exchange or make use of information from another product/system) for Canadian companies abroad.

Clarify the scope of criminal liability

Canada is the only country that has included criminal liability in its AI legislation.

This is a substantial departure from international norms. Bill C-27 should clarify that criminal liability will be for the intentional and egregious use of AI systems for physical harm and serious fraud. Not only will this allow Canada to lead on addressing the use of AI systems for intentional harm, but it will avoid discouraging investment in AI due to the perceived and increased risk of operation.

Remove remote access authority

Minister Champagne’s amendments to Bill C-27 propose a provision that would grant the government broad authority to remotely access the IT systems of companies in and outside of Canada.

This provision raises serious concerns around corporate espionage and cybersecurity. Regardless of the Canadian government’s intent for this legal power, if passed into law, it would set a worrying precedent for governments overseas — particularly those with troubling human rights records.

What does this mean for business?

If Bill C-27 isn’t amended, it will unintentionally limit investment and adoption of AI by businesses of all sizes, further hindering Canada’s already stagnant productivity levels.

Right now, the bill does not provide enough certainty and clarity around definitions, applications or even criminal liability, making it difficult, complex and risky for businesses to integrate AI into their operations.

Since AI — especially generative AI — has the potential to help address Canada’s productivity problem and provide direct business benefits, like accelerating development of creative content and increasing automation in tasks without reducing workforce, the federal government shouldn’t be scrambling to get our AI legislation passed in its current state. Especially when we have the time to get it right for the benefit of all Canadians.

Read the Canadian Chamber’s letter in full here.