Canada's AI Pivot: From Regulation to Rapid Investment

AIDA dies in committee while Carney bets $2.5 billion on hardware

The Abandoned Framework

Legislative Paralysis Meets Technological Momentum

The parliamentary committee room where Bill C-27 died now sits empty, a quiet symbol of policy work left unfinished. Three years of debate, testimony from 130 witnesses, and countless hours of drafting amendments were all suspended indefinitely when Parliament prorogued earlier this year. The Artificial Intelligence and Data Act (AIDA), once positioned as Canada's answer to the EU's AI Act, never made it past committee stage.

Now we have Mark Carney as Prime Minister, and the landscape has shifted dramatically.

The Death of AIDA

AIDA aimed to establish a regulatory framework for the development, deployment, and operation of AI systems, focusing on high-impact systems that could affect health, safety, and individual rights. The legislation faced immediate criticism for its exclusionary consultation process and vague requirements. Civil society groups, labor unions, and creative industries all voiced concerns about its gaps in protection.

Though the government ultimately proposed some amendments in response, they failed to address AIDA's foundational flaws. The House of Commons Standing Committee on Industry and Technology conducted extensive reviews, but progress stalled. While Canada waited for the April 28th, 2025 federal election, the bill was tabled.

The irony? We spent three years debating how to constrain AI while countries like China, which has announced or built over 250 AI data centres, and the UAE, which launched the Stargate compute infrastructure project in partnership with Core42, poured resources into scaling capacity and securing talent.

Enter Carney: The Banker's Approach

During the debate, Chrystia Freeland suggested countering Donald Trump's threat to our borders by rallying like-minded democracies to form a new world order. But Carney went further, identifying AI and cyber capabilities as fundamental to Canadian sovereignty.

"A huge element of securing Canada is going to be related to information warfare, it's going to be related to cyber, it's going to be related to artificial intelligence…It's crucial that we're building those capabilities on Canadian technology, to Canadian values, to protect Canadians, because we can't rely on foreign suppliers for them."

This is not regulatory thinking. It is investment banking logic applied to national infrastructure.

The Liberal platform promised $2.5 billion for digital infrastructure—chips, supercomputing, and data centres—over the next two fiscal years. This builds on Trudeau's 2024 commitment of $2.4 billion over five years but with a crucial distinction: speed. Cohere has already secured $240 million to build a multi-billion dollar AI data centre in Canada, partnering with CoreWeave to bring it online this year.

The Solomon Appointment: A Curious Choice

Perhaps nothing illustrates the new government's approach more than the appointment of Evan Solomon as Canada's first Minister of Artificial Intelligence and Digital Innovation. Solomon was fired by CBC in 2015 after a Toronto Star investigation revealed he brokered art deals with individuals he interviewed, including Carney and Jim Balsillie.

A journalist with ethics concerns leading our AI strategy? The optics alone raise questions. Solomon lacks technical expertise in AI, cybersecurity, or digital infrastructure. That’s a serious gap for a minister expected to evaluate national compute policy, assess high-risk algorithmic systems, and respond to cross-border cyber incidents with strategic precision. His appointment suggests political loyalty outweighed domain knowledge—troubling for a portfolio that demands deep understanding of both technology and its societal impact.

Implications for Standards Work

For those of us working on ISO/IEC JTC 1/SC 27 and SC 42 committees, this shift creates uncertainty. Without AIDA’s regulatory framework, Canadian positions on international AI standards lack domestic legislative backing. We’re contributing to global standards while our own regulatory environment remains undefined.

The Canadian Sovereign AI Compute Strategy allocates:

• $1 billion to fund "public" supercomputing infrastructure

• $700 million for commercial AI data centre construction and expansion

• $300 million to support affordable compute access for small and medium-sized businesses

Infrastructure without governance frameworks is like building highways without traffic laws.

The Real Challenge Ahead

Canada dropped to eighth place from fifth in the Tortoise Global AI Index this year—a decline that signals not just reputational damage but a loss of competitive edge in attracting AI talent, investment, and global partnerships. It now ranks 18th globally for AI infrastructure, behind Germany, France, and the UK. Throwing money at compute capacity addresses symptoms, not causes.

The fundamental question remains: Can we build responsible AI systems without clear regulatory guardrails? The EU has shown that regulation can coexist with innovation. Singapore has demonstrated that sandboxes enable controlled experimentation. Canada appears to be betting on a third path: build first, regulate later.

A Personal Perspective

With experience working at the intersection of technical implementation and policy frameworks, I see both opportunity and risk in Carney’s approach. Compute infrastructure is essential. Accelerating our pace is necessary to stay competitive. But sidelining comprehensive AI governance in favor of pure investment hands the advantage to those who see ethics as a dispensable constraint.

AIDA wasn’t perfect—not by a long shot. But killing it without replacement leaves Canadian organizations navigating AI deployment through a patchwork of outdated privacy laws and voluntary codes. PIPEDA violations carry a C$100,000 fine and extend to AI systems that misuse or improperly collect customer data. That’s a trivial cost for enterprises deploying systems that affect millions.

The appointment of Solomon reinforces the concern. When your AI minister’s main qualification is media experience, not technical competence, you’re prioritizing messaging over substance.

For those of us continuing standards work, we’ll push forward. International collaboration doesn’t pause for domestic political shifts. But without clear national guidance, our contributions lack the credibility they could carry.

Moving Forward

The coming months will test Carney’s gamble. Will infrastructure investment attract talent and firms? Will voluntary frameworks offer sufficient protection? Will international partners see Canada as a leader—or as a regulatory haven?

The concern is not the investment. It’s the abandonment of governance at a time when AI systems are becoming more powerful, opaque, and consequential. The cost of getting this wrong grows exponentially.

Build the infrastructure. Attract the investment. But don’t forget why we pursued regulation in the first place: to ensure AI serves humanity, not the other way around.

The views expressed are personal observations based on public information and do not represent any organizational position.

Afterword: The Compute-First Fallacy

The mathematics in my papers [1][2] demonstrates a simple truth: algorithmic innovation drives exponential gains—a truth Canada continues to overlook as it prioritizes hardware expansion over investing in the very breakthroughs that would make such infrastructure truly transformative. Hardware scaling yields linear returns at best.

Yet we continue to throw billions at compute infrastructure before optimizing our algorithms. Chain of Draft achieved 13× efficiency gains through clever design, not bigger servers. That’s the difference between exp(λA) and (H/H₀)^α.

Canada’s $2.5 billion infrastructure plan exemplifies this backwards thinking. Building data centers before establishing efficient frameworks is like paving highways before inventing wheels.

The path forward? Fund algorithm research first. Optimize frameworks. Then scale infrastructure to match actual needs. Every day we ignore this sequence is a day we waste resources that could fuel real breakthroughs.

Think smarter, not harder.

[1] Think Smarter, Not Harder: Algorithmic Innovation as the Key to Exponential AI Performance
[2] Defining Williams’ Law: The Power of Algorithmic Innovation