Light on specifics, “AI for All” points to a future of limited state impact on the course of AI in Canada — which may be a good thing
Coverage of Prime Minister Carney’s unveiling yesterday of Canada’s long-awaited national AI strategy, “AI for All,” is dominating this morning’s news cycle. You’ve probably already heard the details by now — aiming to build trust in AI, foster literacy, uptake, sovereign compute, and so on.
The document itself , some 50 pages, offers a vivid snapshot of where Canada is at with AI: who’s using it, building it, resisting it, where, when, and how. Well worth reading if you’re curious. In a sentence, Canada is committing to spending $2.3 billion over the next five years, hoping to dramatically increase industry adoption, create a quarter-million new jobs, boost AI development in the private sector, and bolster our domestic infrastructure for AI.
The three critical responses I’ve heard so far are that it’s light on specifics needed to assess whether many of its goals are achievable, such as increasing transparency in AI. It fails to address job-loss risk . And it fails to read the room , urging that we charge ahead with AI adoption with minimal regulation when “an overwhelming majority, 68 per cent, want AI to be regulated heavily, even if it slows down adoption of the technology.” (Even Anthropic this morning is calling , once again, for a slowdown so that “society” can catch up.)
Reading “AI for All,” I made two observations.
The government is bullish on AI, but does not propose any significant new regulatory framework for it.
A new Online Harms Act is coming soon, along with amendments to the Criminal Code on sexual deepfakes , the Elections Act on political deepfakes , and a new consumer-facing privacy bill. But a new privacy law has long been in the works, and the tweaks around deepfakes are important but marginal in the bigger picture. The harm reduction promised in “AI for All” will mainly target children, which we already intended to do in response to social media harms; so now we’ll include chatbots as well.
Similarly, the Strategy touts investments in the Canadian AI Safety Institute to evaluate AI models, and commits to creating a “Canada Trusted AI Certification program to help Canadians identify trustworthy AI products in the marketplace.”
But there is no commitment to passing something like the AI and Data Act tabled in the previous Parliament, a bill that would impose criminal and civil liability for the failure to prevent foreseeable harms. Nor is there a commitment to impose a regulatory framework around the build-out of AI data centres , one that would require an independent audit of the environmental impact of any project the government partially funds.
My second observation is that “AI for All” is filled with good ideas in theory, but many are unlikely to work well in practice.
For example, it aims to provide “all Canadians with access to free AI literacy training, including reaching 1 million entry-level post-secondary students,” and ensuring “all post-secondary students have access to trusted AI agents.” But what would that really entail? To make the training accessible to that many people, it will likely be so simple and watered-down as to be, by now, redundant for many if not most people.
We’re told at one point: “Canada will invest $500 million to expand and enhance the Regional Artificial Intelligence Initiative delivered through Regional Development Agencies to accelerate adoption and commercialization of AI across the country.” And later: “Canada will establish a $500 million Canadian Tech Growth Fund to help close the scale-up capital gap facing Canada’s most promising AI companies. The Fund will provide flexible growth capital and investment support, and enable the federal government, at times, to take equity stakes in the most promising Canadian AI firms.”
I imagine a small business owner in Sarnia, Kamloops, or Rimouski wanting to “accelerate adoption of AI.” Would they really have the time or the inclination to even contemplate how many hoops they would have to jump through to obtain even a dollar in funding from a federal agency for this purpose?
Or take a promising new start-up, founded by a group of recently minted grads from the University of Waterloo. Things are unfolding rapidly; the field is evolving dramatically by the month; competitors are converging on the same business model. More capital would certainly help. And here comes along the government of Canada offering to assist with “investment support.”
Not stated in the document is the reality that to protect public funds from fraud, the procurement process will have to be rigorous — yet somehow quick, flexible, responsive to dynamic conditions on the ground. Seems doubtful the process will be anything but slow and cumbersome, with funding arriving too late or attached to conditions that make it less useful than private capital.
Much of “AI for All” is well intended. But many of its commitments are likely to lead to money spent with little effect.
The strategy as a whole, however, holds out this silver lining: by chasing goals that seem to make sense in light of the current state of AI , it signals the possibility that government will largely stay out of the way of AI’s development in Canada.
It seems far more likely that AI adoption, literacy, trust, and sovereignty won’t emerge from government programs and initiatives, if they emerge at all. They will come instead from the technology itself becoming smaller, lighter, more accessible, more pervasive, and less reliant on the cloud. ■
{To receive new posts by email, follow on Substack .}