Canada’s AI for All Strategy: A Skeptical Guide to the Hype, the Money, and the Missing Guardrails

A Skeptical Guide to the Hype, the Money, and the Missing Guardrails

On June 4, 2026, Prime Minister Mark Carney and AI Minister Evan Solomon launched AI for All, Canada’s new national artificial intelligence strategy.

It is not a small policy memo. It is the federal government trying to reposition AI as economic strategy, industrial policy, workforce plan, digital sovereignty project, public-sector modernization program, and trust-building exercise all at once.

The headline promises are huge: up to 250,000 AI-related or AI-enabled jobs by 2031, up to 90,000 youth jobs and placements, nearly $200 billion in GDP gains, free AI literacy training, a public AI supercomputer, sovereign compute, government procurement for Canadian AI companies, a $500-million Canadian Tech Growth Fund, privacy and online safety legislation, and a Canada Trusted AI Certification program.

That is the optimistic version.

The skeptical version is just as obvious: Canada is trying to accelerate AI adoption before the rules, labour protections, energy plans, procurement pathways, public literacy, and local governance structures are fully built.

So this is not a cheerleading post and it is not a doom thread. It is a follow-along guide for people who want the receipts.

What changed since my first scan

One important correction: the adoption-target mismatch I flagged earlier appears to have been cleaned up. The Prime Minister’s announcement, the full ISED strategy, and the shorter ISED overview now all say Canada wants to raise business AI adoption from about 12 percent to 60 percent by 2034.

That is better than having two official targets on launch day. It does not make the target easy, modest, or well-explained.

The later-day coverage sharpened the story in four places.

First, AP framed Carney’s announcement around foreign dependency and the risk that AI platforms controlled elsewhere could shape Canadian life without Canadian leverage.

Second, Canadian Press quick quotes via CityNews Halifax gathered the predictable but useful split: business groups welcomed the ambition while asking for sharper scale-up policy, labour called for stronger laws and independent oversight, opposition politicians attacked the lack of safety and job-loss detail, Mozilla praised the open-source direction, and News Media Canada pointed out that copyright does not appear in the strategy.

Third, the B.C. infrastructure story got more important, not less. TELUS is pitching its B.C. sovereign AI cluster as 60,000-plus GPUs, 150 MW by 2032, 98 percent renewable power, closed-loop cooling, heat recovery, and a $9-billion economic contribution. But CityNews Vancouver, The Tyee, and the BC Greens show the other half of that story: energy, water, heat, noise, local consent, Indigenous participation, and who gets to say yes before national infrastructure lands in a neighbourhood.

Fourth, the plan is still much stronger on ambition than enforcement. It promises privacy legislation, online safety law, AI model evaluation, and a Trusted AI Certification program. But the public still needs timelines, powers, budgets, independent oversight, appeal paths, worker protections, public procurement rules, and reporting requirements.

What the strategy promises

The official strategy is organized around six pillars:

1. Protecting Canadians and safeguarding democracy.
2. Empowering Canadians through AI literacy and skills.
3. Powering AI adoption across the economy and public services.
4. Building the Canadian sovereign AI foundation.
5. Scaling Canadian AI champions.
6. Building trusted partnerships and global alliances.

The strategy names five priority sectors: health and life sciences, energy and natural resources, transportation, agriculture, and manufacturing and robotics. It also ties AI to defence, dual-use technology, health data, public services, and government procurement.

That matters because this is not just about chatbots in offices. It is about compute, data, procurement, electricity, public services, national security, and who captures the next economic layer.

The government’s own diagnosis is blunt. Canada helped invent modern AI and still has world-class talent, but Canadian businesses have been slow to adopt it. The strategy says only about 12 percent of Canadian businesses used AI to produce goods or deliver services between mid-2024 and mid-2025. It also says Canada ranks near the bottom of peer countries on AI training, literacy, and trust.

That trust gap is the whole ballgame. You cannot order a country to trust AI by press release.

What the coverage is saying

CBC had the pre-release draft story: adoption, literacy, workforce readiness, and a lack of specifics on how harms would be handled. If the direct CBC page is hard to reach, Unpublished’s CBC feed mirror preserves the summary and original link.

Global News went hard on the data-centre and labour questions. Their story notes the plan for large-scale data centres, a world-leading public supercomputer, the 850 MW compute-capacity target by 2030, and the fact that officials did not provide an estimate for how many jobs could be lost through AI adoption.

BetaKit did the startup-policy read: $2.3 billion in spending, few details on privacy regulation, a $500-million Tech Growth Fund, more compute access, national AI institute commercialization funding, and government trying to act as an anchor customer for Canadian AI firms.

The Council of Canadian Innovators has been pushing the “build here or rent forever” line for months. Their submission argues that Canada needs domestic AI champions, Canadian-controlled compute and cloud capacity, and high-value applications powered by proprietary data and IP.

The Logic has been useful on the split inside the policy process. Industry voices pushed for lighter, sector-based or existing-law approaches. Academics and civil society voices wanted stronger rules and clearer protections. That tension did not disappear today. It got wrapped inside the word “trust.”

Vancouver Tech Journal has the local B.C. infrastructure read: the proposed TELUS sovereign AI cluster would span three facilities, scale to more than 60,000 GPUs and 150 MW by 2032, and claims strong sustainability specs.

DIGITAL framed the Web Summit Vancouver roundtable around the move from strategy to adoption: procurement, commercialization pathways, talent, infrastructure, and how Canadian firms actually get bought and scaled.

AP emphasized the sovereignty frame: Carney warned that foreign AI platforms could create risks for Canadian data, firms, and public life.

Canadian Press collected the reaction map. The useful thing is not that everyone reacted according to type. It is that the same unanswered questions kept showing up from different directions: safety, privacy, labour, accountability, open infrastructure, copyright, and whether “trust” has any operational meaning yet.

BC + AI has been making the community-scale argument: national strategy only becomes real when SMEs, creators, educators, nonprofit leaders, public servants, and local builders can turn it into safe workflows, practical training, and accountable adoption.

The money is not neutral

This is not just “AI literacy for everyone.” It is industrial policy.

The government wants a Canadian Tech Growth Fund that can provide capital and sometimes take equity stakes in promising AI companies. It wants to expand the Compute Access Fund by $700 million so SMEs can access sovereign compute. It wants $130 million for commercialization programs across the national AI institutes. It wants to use procurement as a strategic anchor customer. It wants to use the new Sovereign Wealth Fund, where appropriate, to support national champions.

That raises a fair question: who gets picked?

If the state is going to take equity stakes, steer procurement, subsidize compute, and create demand for AI vendors, then the public deserves clear selection criteria.

Which companies qualify? Which regions benefit? What counts as Canadian control? What public-interest conditions attach to public money? How are conflicts managed? What happens if the company is acquired? What happens if the product is unsafe? What happens if the public pays for the infrastructure and private investors capture the upside?

“Canadian champion” can mean a company that keeps talent, IP, and value here. It can also become a polite phrase for public risk and private upside.

The difference is governance.

Trust is not a vibes layer

The strategy says trust is the north star. Good. But trust is not a communications problem. Trust is an operating condition.

Trust looks like privacy law that actually covers AI data extraction and inference. It looks like online safety rules that protect children without turning every platform into a surveillance stack. It looks like deepfake protections that are enforceable. It looks like public registries for government AI use. It looks like independent audits. It looks like incident reporting. It looks like procurement rules that let the public know when an AI system is shaping a decision that affects them.

The announced Canada Trusted AI Certification program could be useful, but only if it has teeth. A trust label without independent testing, appeal paths, incident reporting, and consequences is just a sticker.

The strategy says Canada will work on transparency, including watermarking AI-generated content. Fine. But watermarking is not a governance framework. Watermarks break, bad actors route around them, and the most important harms often come from systems that look boring: hiring screens, benefit triage, insurance pricing, workplace monitoring, automated fraud detection, customer service denials, and predictive policing-style risk scoring.

The scary AI is not always cinematic. Sometimes it is a spreadsheet with an API.

The labour story is thin

The strategy promises jobs. It does not really count losses.

Global News reported that officials were asked whether the federal government had an estimate for job losses from AI adoption and did not answer. Canadian Press captured the same tension in the reaction quotes: labour leaders want stronger laws, independent oversight, protections against surveillance and discrimination, and a greater role for unions in shaping how AI is used.

If Canada is going to push AI into small business, government services, health, agriculture, manufacturing, transportation, and energy, some work will get easier, some work will get better, and some jobs will disappear.

The honest strategy is not “AI creates jobs.” The honest strategy is “AI reorganizes work, and we are choosing who gets support, who gets bargaining power, who gets training, who gets procurement access, and who gets left holding the bag.”

The literacy promise matters. Free training for everyone is better than pretending people can wing it through TikTok tutorials. But literacy cannot be a thin online module that tells workers to prompt better while their employers automate the ladder underneath them.

Real AI literacy includes workflow design, privacy, IP, bias, consent, source checking, documentation, human review, environmental cost, procurement basics, labour rights, and when not to use the tool.

Sovereignty vs. hyperscaler reality

The sovereignty frame is the most interesting part of the strategy and the easiest one to abuse.

The government is right that Canada has a dependency problem. The strategy says Canadian researchers train models on foreign cloud platforms, Canadian companies store sensitive data in foreign jurisdictions, and government operations rely on infrastructure Canada does not own. AP picked up the same theme in Carney’s warning that foreign platforms can be used against Canadians.

So yes: compute matters. Cloud matters. Data residency matters. Procurement matters.

But “sovereign AI” cannot just mean data centres on Canadian soil with foreign chips, foreign model dependencies, foreign cloud contracts, and public money flowing through a small number of incumbents.

Sovereignty should mean Canadian public interest has leverage.

It should mean community benefit agreements. It should mean transparent power allocation. It should mean public-interest compute access for researchers, civic projects, Indigenous data initiatives, artists, and small firms, not only enterprise customers and frontier-model hopefuls. It should mean procurement rules that help Canadian companies without creating a protected insider class. It should mean that data-centre buildout does not quietly become a new extraction economy wearing a maple leaf.

The follow-up question remains the same one I asked in Sovereign AI for Whom?: sovereign for whom?

The B.C. angle

British Columbia is not a side note in this strategy. It is one of the places where the strategy becomes physical.

Web Summit Vancouver gave the federal government a stage to talk about AI ambition in a city with applied-AI companies, creative industries, clean-energy claims, Indigenous data-sovereignty questions, and a growing community of builders trying to do the work without losing their minds or their ethics.

The TELUS B.C. sovereign AI cluster is the clearest example. TELUS says the cluster would include Kamloops, Mount Pleasant, and 150 West Georgia, with more than 60,000 NVIDIA GPUs and 150 MW of capacity by 2032. The company says the facilities will run on 98 percent renewable power, use closed-loop liquid cooling, reduce cooling energy, recover waste heat, and use 90 percent less water than traditional data centres.

If those claims hold up under independent scrutiny, that is better than many alternatives. But better infrastructure still needs public accountability.

The Tyee drilled into the questions people in Vancouver are actually asking: how much power each facility uses, what happens to power rates, whether secured power lets some projects bypass newer competitive allocation rules, how much water is still consumed even with closed-loop cooling, and how the waste heat will be used.

CityNews Vancouver covered the protest side: hundreds marched against proposed AI data centres, citing energy use, water consumption, environmental impacts, and lack of consultation.

The BC Greens are calling for a moratorium on AI data centres until risks are understood and regulated, arguing that local governments are being left to regulate infrastructure they do not have the power or resources to govern properly.

That is the real B.C. test. Not whether we can make a clean-looking data-centre press release. Whether we can build a governance model good enough for the people who live beside the infrastructure.

Who gets access to the compute? What workloads are prioritized? What is the power allocation policy? What are the community benefits? What happens during grid stress? Which host Nations are formal partners? What public reporting exists on water, heat, power, emissions, utilization, public-interest access, and economic benefit?

If B.C. is going to be a demonstration region, then let it demonstrate the whole thing: clean compute, public accountability, Indigenous governance, SME adoption, creator rights, worker training, and local procurement.

The missing copyright and culture piece

One of the sharper Canadian Press reaction quotes came from News Media Canada, which pointed out that copyright does not appear in the 50-page strategy.

That absence matters.

AI is not only a productivity technology. It is also a culture, media, education, and creative-labour technology. It consumes text, images, audio, video, code, journalism, research, public archives, and community knowledge. If the strategy talks about Canadian values but does not explain how Canadian creators, journalists, artists, educators, and publishers are protected or compensated, then “AI for All” is missing a major part of the public.

This matters for small creators as much as big publishers. It matters for Indigenous cultural material. It matters for local journalism. It matters for artists whose work becomes training material without consent. It matters for educators being told to integrate tools that may be built on unlicensed work.

If Canada wants AI literacy, it also needs AI provenance literacy.

What I would watch next

Watch the legislation. The strategy says privacy and online safety laws are coming. The details will decide whether “trust” is real or ornamental.

Watch procurement. If government becomes an anchor customer, the vendor-selection rules matter more than the press release.

Watch the Tech Growth Fund. Equity stakes can align public upside with public risk, but only if the selection and governance are transparent.

Watch the compute deals. Capacity targets sound impressive. The useful questions are who owns it, who controls it, who gets access, at what price, under which public-interest conditions.

Watch the labour numbers. Job creation claims without displacement estimates are half a spreadsheet.

Watch the copyright file. A strategy that says “AI for All” but ignores creators is not actually for all.

Watch the local delivery layer. A national strategy will fail if it does not reach the people who have to implement AI inside real organizations.

My read

Canada needed an AI strategy. Doing nothing would not protect workers, artists, children, small businesses, public services, or Canadian firms. The technology is already here, and the default setting is foreign platforms, private incentives, weak literacy, and scattered adoption.

But AI for All will only mean something if “all” is designed into the machinery.

All means workers, not just employers.

All means creators, not just content pipelines.

All means children and vulnerable people, not just market expansion.

All means SMEs and nonprofits, not just companies already close to procurement.

All means host Nations and local communities, not just land acknowledgments.

All means public-interest uses of compute, not just capacity sold to whoever can pay.

All means the people asking hard questions get treated as part of the build, not as a nuisance outside the room.

There is a version of this strategy worth fighting for. It has clean infrastructure, serious privacy law, practical literacy, worker transition, public-interest procurement, Indigenous governance, creator consent, and Canadian companies that grow here without becoming extractive at home.

There is also a version where “AI for All” becomes a funnel: public money, private upside, weak guardrails, data-centre strain, vague safety promises, copyright silence, and another round of Canadian invention captured somewhere else.

The announcement is not the answer. It is the opening move.

Follow the money. Follow the compute. Follow the procurement. Follow the labour numbers. Follow the water and power. Follow who gets called a champion. Follow whose work becomes training data. Follow who gets called a risk.

That is where the real strategy will show itself.

Follow-along source trail

• Prime Minister’s announcement – launch release, big targets, three guiding principles.
• ISED full strategy – the main document, including pillars, priority sectors, spending lines, compute, skills, and champion-building.
• ISED overview – shorter public summary, now aligned on the 60 percent by 2034 adoption target.
• CBC draft-strategy report – pre-release draft coverage; Unpublished mirror if the CBC page is hard to reach.
• Global News release coverage – strong on data centres, job-loss questions, privacy, public trust, and cloud concentration.
• BetaKit analysis – strong on spending, commercialization, tech-growth fund, compute access, and privacy detail gaps.
• Canadian Council of Innovators submission – scale-up, sovereign compute, proprietary data/IP, and the “build here or rent forever” argument.
• The Logic on the task-force regulation split – useful context on the unresolved safety/regulation debate.
• Vancouver Tech Journal on the TELUS B.C. cluster – local infrastructure details and caveats.
• TELUS release on the B.C. sovereign AI cluster – the company’s sustainability, capacity, and economic claims.
• CityNews Vancouver protest coverage – public pushback on water, energy, environment, and consultation.
• The Tyee’s Vancouver AI data-centre explainer – practical questions on power, water, heat, rates, and regulation.
• BC Greens moratorium call – political pushback on local governance and regulation.
• DIGITAL Web Summit Vancouver roundtable – adoption, procurement, commercialization, and B.C. ecosystem context.
• AP coverage – sovereignty and foreign platform risk framing.
• Canadian Press quick quotes via CityNews Halifax – reactions from business, labour, opposition, open-source, accounting, and media voices.
• BC + AI on trusted local adoption – the regional delivery-layer argument.
• Sovereign AI for Whom? – the earlier data-centre accountability piece this post builds on.