On June 4, 2026, at Toronto General Hospital, Prime Minister Mark Carney announced AI for All, what the government called "Canada's national artificial…

On June 4, 2026, at Toronto General Hospital, Prime Minister Mark Carney announced AI for All, what the government called "Canada's national artificial intelligence strategy." Three pillars (Trust, Opportunity, Sovereignty). Two hundred billion dollars in projected economic growth. A projected 250,000 new AI-related jobs over five years. A target of 60% AI adoption across Canadian businesses by 2034, set against a baseline the strategy's own materials put at roughly 12%, even though Statistics Canada's most recent pre-launch figure was 19.2%. (More on that mismatch below.) And a new international body called the Sovereign Technology Alliance, launched with Germany in February 2026, alongside twelve already-signed international AI partnerships the strategy commits to building on. One note on the adoption baseline. The 12% figure that AI for All uses as the current state is itself the product of rapid recent change. Statistics Canada documented Canadian business AI use rising from 6.1% in Q2 2024 to 12.2% in Q2 2025, a doubling in twelve months, with subsequent StatCan data showing further growth toward the 19.2% rate at the time of AI for All's announcement. Adoption is accelerating. The sectoral pattern matters too: information and cultural industries at 35.6%, professional/scientific/technical services at 31.7%, finance and insurance at 30.6%, all well above the aggregate, with the highest-adoption sectors concentrated in white-collar knowledge work. Among firms already using AI, 25.7% had purchased cloud services or storage and 17.9% had purchased computing power or specialized equipment, confirming that adoption pulls capital toward infrastructure rather than only toward software. The 60% target by 2034 implies roughly a tripling from the current rate over eight years. Whether that target is achievable, what it costs, and whether it's the right target are separate questions the strategy framing doesn't engage. > GOBLIN FACTS — mind the gap between the target and the floor. AI for All aims for 60% business AI adoption by 2034. Statistics Canada's most recent pre-launch reading was 19.2%. That is roughly a tripling in eight years, which is the size of the bet the strategy is quietly making. The next morning, Innovation Minister Evan Solomon repeated and amplified the framing at Mila in Montréal. The morning after that, a coalition called BC + AI (researchers, technologists, Indigenous knowledge holders, public-interest advocates) published a response titled "AI for All Has To Mean All of Us," arguing that the strategy's pillars had to be measured against a different set of pillars including water, power, land, heat, labour, and consent. Three days before any of this happened, the Canadian Broadcasting Corporation had obtained and published a leaked draft of the strategy. The draft had said more than 50% AI adoption by 2030. The final announcement said 60% by 2034 — the number went up, but the deadline pushed four years later. The draft was also blunt about the baseline, citing a KPMG–University of Melbourne study that ranked Canada very low among 47 countries on AI training, literacy, and trust in AI systems. This is what a national AI strategy launch actually looks like in Canada in mid-2026: a hospital announcement, an institutional amplification, a civil society response, a leaked working draft, and a press critique cycle that ran for the rest of the week. The guide's job in this chapter is to give you the architecture to read all of that: to see the strategy clearly, to see the responses to it clearly, and to understand the longer Canadian AI story that the June announcement continues rather than begins. ---

The strategy in plain terms

AI for All has six explicit pillars. The launch day's three thematic banners (Trust, Opportunity, Sovereignty) were unpacked into six working commitments when Minister Solomon formalized the structure in Montréal on June 5, 2026. The guide uses the six throughout:

1. Protecting Canadians and safeguarding democracy (AI safety, deepfake response, online harms) 2. Empowering Canadians (training, AI literacy, transition support — the guide reads this pillar primarily through its labour implications) 3. Powering AI adoption (compute, data centres, uptake across the economy) 4. Building Canadian sovereign AI foundations (Canadian-controlled models, data, infrastructure) 5. Scaling Canadian AI champions (Cohere and other Canadian AI companies) 6. Building trusted partnerships (international cooperation through the Sovereign Technology Alliance)

(Pillar names per the official strategy page, as carried in BC + AI's crosswalk — full citation in the Source Library appendix. The guide's shorthand elsewhere ("Sovereign Foundations," "Trusted partnerships") compresses the official labels.)

The financial commitments named at launch: the AI Compute Access Fund ($300 million at Budget 2024, topped up by a further $700 million at launch to a $1-billion total), with $66 million of the first tranche announced in May 2026 going to 44 small and medium enterprises (including $16.8 million across eight British Columbia projects). The fund's design quietly tilts toward sovereignty: it covers 50 cents on the dollar for compute bought from foreign providers, but 67 cents for compute bought from Canadian ones, a funded nudge toward the sovereign-compute goal rather than a requirement. Over $200 million committed across Mila, Amii, the Vector Institute, and the new Canadian AI Safety Institute (CAISI): $50 million for CAISI, $130 million for AI research commercialization, and funding to raise the Canada CIFAR AI Chairs from 130 to nearly 200 researchers (roughly 70 new chairs). A $100 million Venture Scientist Fund (a private VC fund co-managed by Mila and Inovia Capital, announced alongside the strategy rather than public money) to back AI-native startups incubated at the three national institutes. Up to 90,000 AI-related job opportunities for young Canadians. A commitment to free AI literacy training "for all Canadians" (including reaching one million entry-level post-secondary students and training 3,000-plus educators) delivered partly through the three national institutes.

EXAMPLE — the same loonie, counted five times. Re-announcing money is a politician's favourite magic trick: the funding from 2021 gets a fresh press release in 2024, another in the budget, and a fourth at the ribbon-cutting. Before a billion-dollar headline impresses you, check whether it's a new billion or one old billion wearing four name tags.

The official strategy added a further layer that early coverage compressed: a $500-million Canadian Tech Growth Fund (with Ottawa taking equity stakes in "the most promising Canadian AI firms"), a $500-million expansion of the Regional Artificial Intelligence Initiative, an AI Missions Program opening with $200 million for health outcomes, $100 million each for a Health Sector Data Space and an expansion of the Vital health-data platform, a commitment to a "world-leading" public supercomputer by 2031, a trusted-AI certification program, and Global Talent Stream expansion for AI workers.

Follow the money: every launch commitment in one place.

| Commitment | Amount / note | |---|---| | AI Compute Access Fund | $1B total: $300M (Budget 2024) + $700M at launch. First tranche $66M to 44 SMEs, incl. $16.8M across 8 BC projects. Covers 50¢/dollar for foreign compute, 67¢ for Canadian. | | Institutes (Mila, Amii, Vector) + CAISI | $200M+ total: $50M for CAISI, $130M for commercialization | | Canada CIFAR AI Chairs | 130 → ~200 researchers (~70 new) | | Venture Scientist Fund | $100M (private fund co-managed by Mila and Inovia Capital, not federal money) | | Canadian Tech Growth Fund | $500M (Ottawa takes equity stakes) | | Regional AI Initiative | $500M expansion | | AI Missions Program | $200M, opening with health outcomes | | Health Sector Data Space + Vital platform | $100M each | | Public supercomputer | "world-leading," by 2031 | | AI literacy + youth jobs | up to 90,000 youth opportunities; free training reaching 1M post-secondary students and 3,000+ educators | | Also | trusted-AI certification program; Global Talent Stream expansion |

The "same loonie" caution applies to every row: a headline total can be new money, or old money re-announced with a fresh name tag. Check which before you're impressed.

The framing language to watch (verified in CBC's launch-day report) included Carney's line to reporters, delivered while brushing off questions about US trade friction over the sovereignty push: "This is a strategy any sentient country is taking. This is fundamentally strategic." (Which I'll come back to.) And the stated rationale, in Carney's own words, that Canada "ranks near the bottom of countries in AI training, literacy and trust" — consistent with the KPMG–University of Melbourne study the leaked draft cited.

That's the announcement. Now the analysis.

The first thing to notice is what the strategy commits to and what it doesn't. It commits to spending, to institutions, to international cooperation, and to broad numerical targets (250,000 jobs, 60% adoption, $200B growth). There is no commitment to mandatory disclosure of AI use in environmentally harmful applications. It does not commit to specific Indigenous data sovereignty frameworks or to embedding OCAP, CARE, or NISR principles in federally-funded AI work. Mandatory data-centre PUE or water-use reporting is absent (Canada's voluntary baseline is currently 22% reporting versus 71% globally, per Natural Resources Canada). It does not commit to consent or compensation requirements for AI training data, or to any specific limits on AI use in workplaces or hiring. And it does not name a position on the frontier-risk question Hinton and Bengio have pressed publicly.

This is not a partisan reading. The strategy is what it is: a $300M+ compute fund, $200M+ in institutional support, broad targets, and international cooperation. The list of things it doesn't address is empirical, not editorial. Whether those things should be addressed is exactly the contest the rest of the chapter, and the rest of the guide, is about.

---

The longer story — Canada didn't start in 2026

The federal government acknowledged in November 2025, with the launch of the Treasury Board AI Register, that AI systems have been operating inside the Government of Canada since 1994. That register documented 42 institutions and 400+ AI systems already in use across the federal public service, ranging from research projects to fully deployed tools. The register represents one of two parallel federal AI policy tracks: the AI Strategy for the Federal Public Service, which governs how government internally uses AI, distinct from AI for All, the broader Canada-wide economic and societal strategy. These tracks are often collapsed in coverage. The guide keeps them separate.

The Pan-Canadian Artificial Intelligence Strategy launched in 2017 — the world's first national AI strategy. It established the Canada CIFAR AI Chairs program, the three national institutes (Mila in Montréal, the Vector Institute in Toronto, Amii in Edmonton), and the broad architecture that AI for All now extends. Mila, founded by Yoshua Bengio (one of the three 2018 Turing Award recipients for foundational work on deep learning); Vector, founded with University of Toronto affiliations including Geoffrey Hinton (another of the three Turing recipients, and later a Nobel laureate); Amii, originating in 2002 as the Alberta Ingenuity Centre for Machine Learning with University of Alberta ties.

What this means in practical terms: the technology underneath ChatGPT was developed in significant part by Canadian researchers, in Canadian institutions, with substantial Canadian federal funding, over a period of more than two decades before it became a consumer product in late 2022. Hinton, Bengio, and Yann LeCun's joint 2018 Turing Award was given for the deep learning architectures (convolutional neural networks, backpropagation methods, attention mechanisms) that the current generation of large language models is built on. The story that Canada is "behind on AI" needs to be qualified by the story that Canada was foundational to AI. Both are true, and they explain different things.

The Pan-Canadian strategy was renewed in 2021 with $443 million in federal commitment. The 2024 federal budget allocated $2.4 billion for new computing infrastructure and safety research. In December 2024, the federal government committed $240 million to Toronto-based Cohere for sovereign compute capacity. That money backs a Cambridge, Ontario facility built and operated by the US cloud provider CoreWeave, with Cohere as tenant. Bell's separate AI Fabric project is a roughly 500 MW build announced in May 2025; Bell and Cohere partnered on it that July. It became one of the two highest-profile Canadian AI infrastructure announcements before AI for All. The other was the O'Leary/Greenview Wonder Valley project in Treaty 8 territory, discussed in Chapter 6, including the consultation dispute now before the courts.

The Artificial Intelligence and Data Act (AIDA), introduced as Part 3 of Bill C-27, was the previous Trudeau government's attempt at a comprehensive AI regulatory framework. AIDA died when Parliament was prorogued in January 2025. Minister Solomon has said publicly, in a June 2025 interview with The Logic, that he will not reintroduce AIDA in its previous form. Some elements may be carried forward in future legislation. The other major AI-adjacent bill from the previous Parliament, the Online Harms Act, also died. Its successor, the Protecting Victims Act (Bill C-16, December 2025), covers only sexual deepfakes, not the broader online-harms regime AIDA's parent bill had aimed at.

Between the failed comprehensive legislation and AI for All, the federal government also pursued a voluntary-framework path. In September 2023, ISED launched the Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems, signed at launch by a named set of Canadian and Canada-operating AI actors (Ada, the Toronto customer-service company; BlackBerry; Cohere; Coveo; OpenText; and TELUS among the originals) and joined in later waves by CGI and IBM (December 2023), then Mastercard and Salesforce (May 2024), among others. Microsoft, notably, had not signed as of June 2026, a fact worth holding onto when weighing what voluntary frameworks capture and what they don't. (Receipt: the ISED signatory page, checked June 2026, lists 46 organizations, including, pleasingly, Amii, Mila, and the Vector Institute themselves. Microsoft remains absent.) The Code commits signatories to accountability, safety, fairness, transparency, human oversight, and validity-and-robustness standards in the development and deployment of advanced generative AI. It is voluntary, with no enforcement mechanism, no penalty for non-compliance, and no mandatory disclosure, but the signatory list itself is informative. The Code is one of the documented engagements between Canadian AI industry actors and federal AI governance, sitting in the gap between failed comprehensive legislation (AIDA) and the investment-strategy framing of AI for All. Chapter 17 returns to the Code as part of the broader Canadian AI governance landscape.

The consultation that informed AI for All itself produced substantial civic engagement. The federal consultation drew over 11,300 public submissions and 32 expert reports between its launch in 2025 and the consultation summary published in February 2026. That participation scale is meaningful — public engagement on Canadian AI policy is not confined to industry insiders or specialized stakeholders. Whether the consultation's outputs were proportionately reflected in AI for All's final commitments is a separate question; the consultation summary is publicly available for readers who want to compare submissions with adopted strategy elements.

So the Canadian AI story arrives at June 2026 with: thirty years of federal AI use, almost a decade of explicit national AI strategy, foundational global research credentials, two failed major regulatory bills, a new compute infrastructure push, a parallel public-service AI strategy, and a still-unresolved set of debates about what to do legislatively. AI for All is the latest chapter, not the first.

---

The leaked draft, and what it tells us

On June 1, 2026 (three days before the AI for All announcement), the CBC published a draft of the strategy obtained from inside the government. The reporters, Lopez Steven and McKenna, were straightforward about what they had and what it told them.

Three things in the leaked draft are worth holding onto.

One: the target shifted. The draft set more-than-50% Canadian business AI adoption by 2030. The final announcement set 60% by 2034. The number went up, which looks more ambitious, while the deadline slid back four years, which makes the timeline less aggressive. Either move alone would be a political signal. Both at once is worth a note.

Two: the draft was unusually candid about the structural problem the strategy is meant to address. It stated that Canadian small and medium enterprises "have no affordable domestic option" for AI compute and that they "train and deploy models on foreign cloud platforms," which "sends Canadian money abroad" and "places sensitive data and intellectual property outside the country." That sentence is exactly the rationale for the compute fund and the sovereign AI infrastructure investment, and it doesn't appear in the public release in that direct form. The internal candour was sanded down. Or, to say only what the documents show: the sentence exists in the draft and not in the release. The sanding is an inference. The diff is not.

Three: the draft was candid about Canada's standing. It cited a KPMG–University of Melbourne global trust study finding Canada ranked very low among 47 countries on AI training and literacy and on trust in AI systems — the evidence base beneath the launch's literacy-and-trust framing. The candour about the diagnosis survived into the announcement; the unflattering international ranking mostly didn't.

(And a withdrawal, because the methodology applies to the guide: an earlier draft of this chapter attributed to the CBC story a cross-party thread on blocking superintelligent AI development. The CBC article contains no such thread. Until a source surfaces, the guide treats that claim as unverified and builds nothing on it. Receipts or it didn't happen.)

What the leak shows, in plain terms, is the gap between internal working positions and external presentation. That gap exists in every government, on every file. Naming it specifically here does two things: it lets readers calibrate their reading of AI for All (the announcement is the polished version of a more conflicted internal conversation), and it shows where the most informative questions for journalists, opposition parties, and civil society to ask actually are: not in the press release, but in the gap.

---

The civil society response

On the same day as Solomon's Montréal amplification, June 5, the BC + AI coalition published "AI for All Has To Mean All of Us," written by Kris Krüg, with contributions from a network of researchers, technologists, Indigenous knowledge holders, and public-interest advocates based primarily in British Columbia. (Disclosure, because the guide's methodology has to apply to the guide: the author is a member of this community, and Krüg is a friend and mentor. The lean toward taking this response seriously is real. Now it's labelled.) The response was not a rejection of the strategy. It was something more useful: a 6-pillar crosswalk mapping each federal pillar to a regional, public-interest delivery test, asking what it would take for the pillar to reach the people who "will live with the systems, feed the datasets, pay the power bills, defend the water, protect the culture."

The published crosswalk, compressed:

| Federal pillar | BC + AI's delivery test | |---|---| | Protecting Canadians and safeguarding democracy | Fund public-interest governance capacity — community review, plain-language risk literacy, procurement rules people can actually contest | | Empowering Canadians | Practical AI literacy through trusted local intermediaries: creators, workers, educators, seniors, newcomers, Indigenous communities | | Powering AI adoption | Move beyond tool demos to cohort-based adoption, "with labour impacts in the room from day one" | | Building Canadian sovereign AI foundations | "Tie compute, data, power, water, land use, Indigenous governance, and climate accountability together before the infrastructure is locked in" | | Scaling Canadian AI champions | Use public demand to grow companies whose systems are "auditable, rights-respecting, accessible" | | Building trusted partnerships | Treat regions, civil society, creators, labour, and Indigenous governments "as delivery partners, not as atmosphere around a federal strategy" |

The post's sharper questions sit in its prose rather than its table — "If compute is sovereign, sovereign for whom?"; "who benefits, who governs, who gets access, and which communities carry the costs?"; data centres as "climate questions wearing a hoodie." The guide's own gloss, for what the crosswalk implies and the strategy doesn't answer: worker consent before deployment rather than retraining after; water, power, land, and heat counted alongside megawatts; and whether alliance-with-the-hegemon inside the Sovereign Technology Alliance is what "sovereignty" means. That gloss is the guide's; the receipt for what BC + AI actually published is in Sources/.

The single line to carry forward, because it compresses so much: "Sovereignty is not achieved by putting a maple leaf on the invoice." When the rest of the guide reaches the question of what "sovereign AI" means in practice (Chapter 9), this is the line to test claims against. A Canadian-flagged subsidiary of a US hyperscaler is not the same as a Canadian-controlled AI infrastructure. The invoice can say Toronto while the encryption keys, the contract law, the data governance, and the underlying chip supply remain elsewhere.

🧌 GOBLIN CHECK — "Sovereignty is not achieved by putting a maple leaf on the invoice" is the best single sentence of launch week, and it gets a permanent shelf in the goblin's hoard. Every time someone says "sovereign AI," check three things: whose law, whose chips, whose capital. A flag on the contract answers one of the three. On a good day.

A second BC + AI finding is worth noting. Citing Statistics Canada's May 27, 2026 data, they surfaced that 19.2% of Canadian businesses were using AI as of that release, up from 12.2% in the second quarter of 2025. The strategy's headline target is 60% by 2034. The current baseline is 19%. That's a tripling over about eight years, and it implies either rapid uptake the underlying ecosystem isn't currently structured to support, or numerical inflation in the target.

A third observation, the guide's own, builds from where the BC + AI response points and connects to the Indigenous data sovereignty conversation Chapter 9 deepens. The federal AI for All press release explicitly names consultations with "industry, academia, unions, communities" and beneficiaries including "workers, entrepreneurs, researchers, students, and community leaders." It does not name Indigenous Peoples or Indigenous communities as a consultation pillar. By contrast, the Carney government's Major Projects Office established an Indigenous Advisory Council in September 2025 (11 members, formal mandate). So the absence isn't because the government doesn't know how to structure parallel Indigenous consultation processes. The omission from AI for All is its own data point.

---

The press critique

Mainstream press coverage of AI for All fell into roughly three tones, useful to distinguish.

The neutral-but-pointed. The Globe and Mail (Tom Cardoso and others) ran a piece foregrounding IP lawyer Jim Hinton's critique that the strategy "was crafted as if it was done in a vacuum, forgetting that the U.S. and China are the primary owners and controllers of AI technology." The Globe also flagged that the launch occurred at Toronto General Hospital, a health-coded setting that emphasized AI-enables-services framing over AI-displaces-workers framing. The Hinton quote is the sharpest single-sentence critique to come out of launch week, and it tests directly against BC + AI's "maple leaf on the invoice" line.

The internationally skeptical. Digital Journal covered the U.S. Trade Representative's 2026 National Trade Estimate Report, which flagged Canada's Sovereign Cloud Initiative as a non-tariff procurement barrier. That's a significant finding: the country AI for All names as Canada's most important trading partner is officially treating elements of the sovereignty strategy as protectionism that may require trade-policy response. The implication for the strategy's "Trusted partnerships" pillar is uncomfortable.

The headline-skeptical. Engadget ran coverage describing the strategy as "questionable" in the headline itself, unusual editorial framing for a national policy launch. The substance underneath was a milder version of the Globe and BC + AI critiques, but the headline registered a tone worth noting: international tech press did not receive the strategy as a clear success.

CBC's launch report carried Carney's line to reporters, delivered when he was pressed on whether the sovereignty push would aggravate US trade relations: "This is a strategy any sentient country is taking. This is fundamentally strategic." Precision matters here. The line wasn't a general boast about the whole strategy; it was a defence of the sovereignty pillar specifically, framing sovereign compute as so self-evidently necessary that questioning it approached questioning sentience. The response cycle showed it was neither obvious nor unanimous. The same article carried the NDP calling the strategy "reckless and inadequate" ("Now they're mouthing the lines of tech billionaires," said MP Don Davies) and Conservative deputy leader Melissa Lantsman doubting the jobs math against already-high youth unemployment. The literacy-and-trust admission has its receipt too: Carney himself noted Canada "ranks near the bottom of countries in AI training, literacy and trust," consistent with the KPMG–University of Melbourne study the leaked draft cited. It implicitly accepts that the strategy's "Empowering Canadians" and "Building trusted partnerships" pillars are being built on a low baseline.

The fair summary of the press cycle: the strategy was treated as a substantive policy event, took meaningful critique on its first day, and did not generate the "Canada is now competitive globally" framing the government appeared to be reaching for. That's neither a triumph nor a catastrophe. It's the actual reception, which is worth seeing clearly because both critics and supporters tend to misremember the launch as the version that confirms their priors.

---

The two cabinet voices to track

Two ministerial voices have done most of the public-facing work on AI for All, and the guide will track both throughout.

Evan Solomon is the Minister of Artificial Intelligence and Digital Innovation. His Montréal amplification on June 5 (at Mila) formalized the six-pillar structure of the strategy. He has been the more pragmatic, business-facing voice, comfortable with the language of Canadian champions, compute, commercialization, and international partnership. When Grok was used to generate non-consensual sexual deepfake images in January 2026 (the case we'll come back to in Chapter 13), Solomon's response pointed to the Protecting Victims Act (Bill C-16) rather than to broader AI regulation. He has publicly stated he will not reintroduce AIDA in its previous form.

Mark Carney as Prime Minister has been the more sweeping voice: the "any sentient country" framing, the Toronto General Hospital launch staging, the explicit pitching of AI as economic transformation. Carney's own background (Bank of Canada and Bank of England governor, climate-finance work at GFANZ, a global net-zero finance coalition) brings a financial-system framing to the AI conversation that's distinct from the technology-policy framing AI debates usually have. The implication, that Canadian AI policy is being shaped by someone who thinks about it primarily through the lens of macroeconomic competitiveness and capital allocation, is worth holding through the rest of the guide.

Two other names worth filing: Shafqat Ali, President of the Treasury Board, launched the federal AI Register in November 2025 (the parallel internal-use track distinct from AI for All). Sean Fraser as Justice Minister tabled Bill C-16 in December 2025. These are the cabinet members whose specific portfolios have meaningful AI components beyond Solomon's direct mandate.

---

The honest assessment

It is possible to hold all of the following simultaneously, and the guide will spend the rest of its chapters showing why each one is supportable:

AI for All is a substantive policy event. It commits real money, builds real institutions, signals real international positioning, and represents Canada's most ambitious AI policy intervention to date.

AI for All does not commit to several things that civil society, academia, labour, and Indigenous voices have spent years arguing should be central to AI policy. The silences are not random. They cluster around: mandatory transparency, Indigenous data sovereignty, worker consent (as distinct from worker training), creator/IP rights, environmental disclosure, and the foundational-risk question.

AI for All builds on three decades of Canadian AI capacity that the strategy doesn't always foreground — partly because acknowledging the existing capacity complicates the "Canada is behind" framing the strategy uses to justify its ambition. The Treasury Board Register's 1994 anchor is the most direct rebuttal of that framing, and it was published by the same government, six months before AI for All.

The AI for All numbers (250,000 jobs, 60% adoption, $200B growth) are projections, not commitments. Projections in government strategy documents are made by people, with assumptions, on schedules. The guide will treat them as projections throughout, with the underlying assumptions named where possible.

The AI for All response cycle (leaked draft, civil society crosswalk, mainstream press critique, international trade pushback) is itself the substantive Canadian AI conversation. The strategy didn't end the debate; it gave the debate its current shape. Reading the response cycle alongside the strategy is how you actually understand what Canadian AI policy is.

---

Where the story stands

CHAPTER RECAP — you now have: - The AI for All launch read structurally — three thematic banners (Trust, Opportunity, Sovereignty) unpacked into six operational pillars, with the gap between what the strategy commits to (spending, institutions, international cooperation) and what it leaves voluntary held in view. - The longer story behind the launch: the Pan-Canadian AI Strategy since 2017, the failure of AIDA (Bill C-27), and the September 2023 Voluntary Code of Conduct that filled the legislative gap with non-binding commitments. - The target shift caught on the record — the leaked draft's roughly-50%-business-adoption-by-2030 became 60%-by-2034 in the final announcement: the number went up while the deadline slid back four years. - The five-document method for reading any AI announcement: the launch is one document; the leaked draft, the civil society crosswalk (BC + AI), the press critique, and the international response together form the actual policy event. - The two voices doing the public-facing work — Evan Solomon (Minister of AI and Digital Innovation) and Prime Minister Mark Carney — and what each signals about where the strategy is heading. - The financial and geopolitical scaffolding under the "sovereignty" banner: the $500-million Canadian Tech Growth Fund and the Sovereign Technology Alliance (launched with Germany, February 2026).

The next chapter (Chapter 6) takes the physical layer of the strategy's "Powering AI adoption" pillar and goes inside it — the actual data centres, the actual megawatts, the actual water, the actual sites (Wonder Valley, Calgary, Bell AI Fabric's footprint). The hyperscaler landscape (AWS Canada Central, Azure Canada, Google Cloud Toronto and Montréal) gets mapped against the "sovereign AI" framing to see what the framing actually claims and what it doesn't.

You now have the architecture for reading any future Canadian AI announcement, and any pushback on it, in the right structural way. The announcement is one document. The leaked draft, the civil society crosswalk, the press critique, and the international response form the actual policy event. Read all five and you'll know what's actually happening. Read only the announcement, and you'll know what the government wanted you to think is happening.

ALIGNMENT — the five-document rule. Every strategy launch is really five documents: the announcement, the leaked draft, the civil-society response, the press critique, and the international reaction. Read only the press release and you know what the government wished you knew. The distance between the draft and the launch is usually the most honest page in the pile.

---

Bias label for this chapter: structural-political analysis of a current government policy launch. Author lean: critical-engaged, sympathetic to civil society and Indigenous critiques surfaced during the launch cycle, skeptical of strategy framing that doesn't match operational commitments. Sources labelled inline where they appear. Government voices (PMO, Solomon, ISED) treated as primary on the strategy itself and as leaning toward justification of the strategy. Civil society voices (BC + AI), press critique (Globe, Engadget, Digital Journal, CBC), and leaked working documents treated as primary on the response cycle.

Primary sources cited or relied on in this chapter: PMO June 4, 2026 launch announcement; ISED May 12, 2026 compute fund release; Minister Solomon June 5, 2026 Montréal amplification; Treasury Board AI Register (November 28, 2025); CBC leaked draft coverage (Lopez Steven & McKenna, June 1, 2026); BC + AI "AI for All Has To Mean All of Us" (Krüg, June 5, 2026); Globe and Mail launch coverage including Jim Hinton; Digital Journal on USTR 2026 National Trade Estimate Report; Engadget launch coverage; Statistics Canada AI adoption analyses for Q2 2024 and Q2 2025 including sectoral breakdown; ISED Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems (September 2023) with signatory list; Federal AI Strategy consultation summary (February 2026) documenting 11,300+ submissions and 32 expert reports. Detailed citations in the Sources appendix.

---