What Canada's National AI for All strategy means for trust, jobs, and excluded populations

Energy implications

The strategy commits to AI data centres that "scale to at least 100 megawatts," to 850 MW of compute capacity by 2030 and scaling capacity "of up to 2.3 GW with corresponding investments in the tens of billions." It pairs this with grid context: "more than 83 percent of the country's electricity grid comes from renewable and low-emission sources," and data centres on those sources "reduce their total operating emissions by up to 90 percent."

At continuous draw, 850 MW is about 7.4 TWh per year (850 MW × 8,760 hours), and the 2.3 GW scaling figure is about 20 TWh per year. Against Canada's 625.7 TWh of electricity generation in 2021 (Canada Energy Regulator), that is roughly 1.2 percent and 3.2 percent of national generation; against the residential sector's 177 TWh in 2020, roughly 4 percent and 11 percent. The draw is material at national scale, and the strategy pairs it with grid expansion rather than drawing it from existing supply, which is why the physiological burden below is scored small and capped.

The strategy ties new data centre development to "clean energy expansion, robust environmental standards, and tangible benefits for local communities." The commitment is stated; the strategy does not define how a "tangible benefit for local communities" is measured. That missing methodology is the subject of the Foundation's anchor-tenant audit, which scores recovered heat, on-site community infrastructure, and net energy burden at the point of delivery. Without a community-benefit metric, the commitment cannot be audited or compared across sites.

Sovereignty implications

Pillar 4 builds "the Canadian sovereign AI foundation." The strategy states that "Canada's sovereign compute capacity is nascent, leaving Canadian organizations reliant on foreign providers," and responds with $700 million in "affordable sovereign compute through an expansion of the Compute Access Fund," a "world-leading public supercomputer…by 2031," and the Sovereign Technology Alliance that Canada and Germany launched in February 2026.

Sovereignty over compute is fulfillment the strategy emits: it reduces a named dependency. The framework distinguishes infrastructural sovereignty (who owns the machines) from value sovereignty (whose needs and values the machines are trained on). The strategy treats the first directly and the second through one channel, examined below.

Labour implications

The labour targets are stated as job creation: "up to 90,000 AI-related jobs and work placement opportunities…by 2031" and "up to 250,000 new jobs through the adoption of AI by 2031," alongside the productivity target of "a 3% increase in GDP, representing nearly $200 billion in GDP gains." AI literacy is the delivery mechanism: content that will "reach 1 million entry level post-secondary students and train more than 3,000 educators with AI learning kits."

These are gross-creation and throughput figures. The strategy states no displacement target: it names no figure for jobs lost, eliminated, or transitioned as adoption reaches 60 percent. It commits to "track and assess the societal, labour market, and economic impacts of AI" through Statistics Canada's measurement program, without publishing the metric by which a net labour effect, after displacement, would be reported. The framework scores the strategy on what it absorbs and emits, so the gross job creation is credited as emitted fulfillment and the unmeasured displacement risk is the reason the safety level absorbs no deprivation.

The displacement is not hypothetical. The International Monetary Fund estimates that almost 40 percent of global employment is exposed to AI, rising to about 60 percent in advanced economies such as Canada, and that for roughly half of exposed jobs AI "may execute key tasks currently performed by humans," lowering labour demand and wages (IMF, January 2024). A strategy that targets 60 percent business adoption while reporting only gross job creation measures one side of that exposure.

The Foundation's economic case sets out where the demand goes. It cites Metaculus community forecasts of US labour by 2035: employment 3.4 percent below the Bureau of Labor Statistics baseline, the most AI-exposed occupations down 17.2 percent, labour's share of national income falling from 62 to 56 percent, and new four-year-graduate unemployment doubling from 6 to 12 percent. The shrinking work is "desk-bound symbolic-manipulation jobs that AI can substitute for"; the growing work is "high-touch, in-person, emotionally and relationally loaded: registered nurses, K-12 teachers, restaurant servers." That relational labour is an unmeasured wedge the Foundation estimates at $11 trillion per year or more. What gets measured gets a market: without measurement infrastructure for relational work, no pricing or compensation grade can form for the jobs the same forecasts say will grow. The strategy's targets measure adoption, literacy, and gross job creation; it defines no metric for the relational labour these forecasts place on the growing side.

Community, place, and delivery channels

The strategy routes AI literacy through existing civic infrastructure: it will "empower public libraries and community organizations, long trusted as hubs for learning, as natural partners to bring AI literacy initiatives into every community," "with emphasis on reaching underrepresented groups" and "those in rural, remote, and northern regions." It commits to French-language tools "in Québec and in Canada's vibrant Francophone minority communities," and to applying "Gender-Based Analysis Plus in a meaningful way across policy design."

These channels reach populations often excluded from the AI economy, and they reach them where they already gather. The flow runs one direction: AI literacy, tools, and access are delivered to communities. The strategy's promise that Canadians "of every background" have "a meaningful voice in the public debates that will define AI's role in Canadian society" is participation in debate, distinct from sourcing the values that train the systems. The one section where the flow reverses is examined next.

Ethical implications

Trust is named as the strategy's binding constraint: "Trust is the north star of this strategy," set against the ranking of 42nd of 47 on trust in AI systems. The mechanisms are protective and capability-building: Pillar 1 on safeguarding democracy, $50 million to expand the Canadian AI Safety Institute, and Gender-Based Analysis Plus across policy design.

One section sources values from a community rather than delivering capability to it. The strategy commits to "support Indigenous self-determination over how AI is built and used in Indigenous contexts," naming the Indigenous Languages Program at Canadian Heritage, the Indigenous Languages Technology Program at the National Research Council, and First Languages AI Reality at Mila. This is the template the Foundation's research argues for across other excluded populations: communities as a source of values for AI.

The strategy extends this values-sourcing approach to no other population. The Foundation's study Whose Values Train AI? found that 43 percent of 51 interviewed Downtown Eastside residents named compassion, care, or kindness as the primary thing AI should learn from people, a category absent from the 81,000-person Anthropic survey of what people want from AI. For every excluded population beyond Indigenous communities, the strategy's mechanisms are literacy, tools, and access. None treats those populations as a source of the values AI is trained on. That is the gap this evaluation measures.

Where the strategy sits on the quadrant

The Transmutarianism framework scores agents on F (filtering of deprivation: deprivation absorbed without being passed on) and A (amplification of fulfillment: fulfillment emitted in excess of what was received). Moral work M = τF + A is computed per Maslow level (physiological, safety, belonging, esteem, actualization) and weighted by w = {5, 4, 3, 2, 1}.

The chart below plots the strategy's provisional placement (sigil-red outline) against the four archetype reference dots (muted, shown for orientation). Drag the τ slider to test sensitivity. Click any dot for its F, A, and M values.

Assumptions, stated

1. Scope of evaluation. The published National AI for All strategy as of 2026-06-04. Subsequent budget instruments, departmental implementation plans, and provincial strategies are excluded. The strategy is scored as written, on its stated targets and named programs.
2. Time horizon. Steady-state at the strategy's own target years (2031 for jobs and the public supercomputer, 2034 for adoption). Implementation-phase activity is real but is not scored against the steady-state dot.
3. Asymmetry coefficient (τ). Default τ = 1. Sensitivity at τ = 0.8 (flourishing focus) and τ = 1.5 (cycle-breaking focus) is shown in the math box below.
4. Maslow weighting. w = {5, 4, 3, 2, 1} for {physiological, safety, belonging, esteem, actualization}. Lower-level deprivation is weighted more heavily.
5. F and A scale. Each level scored on a −10 to +10 range, central estimate from the public-record evidence above. The reasoning column names the inputs.
6. Denominator. The primary dot counts flows to all Canadians, the strategy's own denominator. A second view, reported in the sensitivity band, counts flows to populations excluded from AI alignment (low-income, rural, remote, and northern, Indigenous, low-literacy, and Downtown-Eastside-type populations).
7. Treatment of throughput. The primary dot credits only flows that demonstrably reach human need; raw jobs, adoption, and GDP figures are scored near zero as amplification, because dollars moved and positions created are not, by themselves, deprivation transformed. A second view, reported in the sensitivity band, credits the throughput targets at face value. This is the single most placement-determining choice in the evaluation.

Per-Maslow scoring

The aggregate dot above is computed from the per-level table below, on the primary view (all Canadians, throughput discounted). Each F and A value is the central estimate from the public-record evidence in the prior sections; substitute different numbers and the dot moves.

The placement is insensitive to τ, because F is near zero: changing τ scales the F contribution, and the F contribution is small, so W barely moves across the [0.8, 1.5] band. The result is governed almost entirely by emission; the absorption of deprivation contributes close to nothing. The two assumptions that do move the dot are the denominator and the treatment of throughput. Under an excluded-populations denominator the dot crosses the F axis into Magnifier, because most of the emitted fulfillment reaches businesses, students, and the tech sector rather than the most excluded; the Indigenous-led section is the part that holds. The Foundation will publish a re-audited placement against any party that supplies population-resolved flow data.

Public-interest recommendations

The strategy names trust as its north star and commits to measuring the societal impacts of AI. The recommendations below address the metric that commitment still needs.

For Innovation, Science and Economic Development Canada

1. Publish a deprivation-transformation metric alongside the throughput targets: for the 90,000 and 250,000 job targets, report the net effect after displacement, by region and by population.
2. Define "tangible benefits for local communities" for data centres as measurable quantities (recovered heat in GWh per year, on-site community infrastructure, net energy burden), so the commitment is auditable across sites.
3. Extend the Indigenous self-determination model to other excluded populations as a values-sourcing mechanism, not only a literacy-delivery one.

For Statistics Canada

1. Within the Artificial Intelligence and Technology Measurement Program, specify the societal metric. Pair the labour-market and economic series with a measure of how AI changes deprivation for the populations least represented in adoption data.

For provinces and utilities

1. Attach community-benefit measurement to data centre siting and power agreements before approval, so the federal "tangible benefits" commitment has a provincial point of delivery.

For libraries, community organizations, and Indigenous partners

1. Record what communities say AI should learn from them, not only what literacy they receive, and supply that record as the values input the strategy currently sources from one community.

The audit protocol (six inputs)

An audited placement of the strategy, or of a program within it, requires six inputs. The Foundation offers framework, scoring template, methodological support, and the published report at no cost to any party willing to supply them.

1. Program allocations by instrument, with committed amounts and conditions.
2. Target outcomes stated as flows reaching named populations, not only as national aggregates.
3. For compute and data centres: capacity, energy source, net energy burden, and community-benefit infrastructure measured at the point of delivery.
4. Labour outcomes net of displacement, by region and by population.
5. The societal metric the Statistics Canada program will report, with its definition.
6. The values-sourcing record: what excluded populations state AI should learn from them, by community.

Contact: sev@economyofwisdom.com.

What changes the placement

Toward Transmuter: a published net-of-displacement labour metric; a measurable definition of data centre community benefit; a values-sourcing mechanism that extends the Indigenous self-determination model to other excluded populations; a societal metric in the Statistics Canada program that reports deprivation change for under-represented groups. Each of these raises F, the dimension currently near zero.

Toward Extractor: if the throughput targets are met while the societal metric remains undefined, if data centre benefits are claimed without measurement, and if adoption reaches 60 percent with displacement unreported, then emitted fulfillment concentrates and absorbed deprivation stays at zero or turns negative, moving the dot down and left.

Better data moves the dot.

A note on framing

The strategy and the Foundation share premises. Both name trust as the binding constraint rather than compute or talent. Both treat Indigenous-led AI as a source of values rather than a recipient of tools. The divergence is in mechanism. A government operates within the measurement systems it inherits, and incentive structures outlast the intentions of the people inside them. You do not change such a system from inside it. You change it by measuring its systemic effects, if the measurement tool earns enough attention to matter.

This field note scores a strategy that measures its success in throughput: jobs created, dollars moved, points of GDP. The framework asks a different question, of the strategy and of the Foundation alike: how much deprivation was transformed, and for whom. The placement here is an instrument for asking it, contestable on every assumption stated above.

The math is at transmutarianism.org/framework/. The live quadrant explorer is at transmutarianism.org/quadrant/. To dispute the placement, substitute different F and A values per level, with a source for each, and the dot moves.