Today Mark Carney launched the Prime Minister’s National Council on Artificial Intelligence: thirteen people tasked with advising the government’s “AI for All” strategy. It is a serious list, and the mandate is serious—sovereign AI infrastructure, economic growth, adoption and safety.
Read the list once more. Then ask the question I asked: who among them is there for the data?
The missing seat.
A model is its training data. Everything this council will advise on—adoption, competitiveness, safety, sovereignty—is downstream of one question: what gets to train the models, and who decides?
The council has builders, researchers, investors and governance experts. It does not have a single person whose job is the data supply layer: finding corpora, clearing rights and moving training data to builders at speed.
In the United States, that industry exists—Mercor, Micro1, Scale and a dozen others. In Canada, the role does not exist on anyone’s org chart, let alone on the Prime Minister’s council.
A model is its training data. The supply chain needs a seat.
Sovereignty is decided at the data layer.
“AI for All” promises sovereign AI infrastructure. But infrastructure is not sovereignty.
Canada already ran this experiment: C$2.4 billion for AI advantage, including $2 billion for compute—the AI Compute Access Fund and the Sovereign Compute Strategy. The data layer got no comparable mission.
A data centre training on corpora licensed from American aggregators, under American contracts, is not sovereign AI. It is a building with a flag on it.
Now the council will spend its energy on adoption targets and job numbers while the actual territory gets divided. Every month, another Canadian corpus—news archives, industry records, bilingual corpora, operational data from the resource industries—gets signed into a foreign exclusive. None of it comes back.
Builders can build only with what someone has sourced.
Data is not a passive input waiting at the end of a policy pipeline. It has to be discovered, rights-cleared, documented, governed and moved. That is an operating function—and a strategic one.
What the council should do.
Three things, and none of them are complicated:
- Put a data-supply voice at the table. Not a researcher who studies data. Not a lawyer who regulates it. Someone who sources it, clears it and moves it. The supply chain has a seat in every other strategic industry. AI should be no different.
- Treat corpus infrastructure like compute infrastructure. A fraction of the Sovereign Compute Strategy aimed at corpus discovery, rights clearance and licensing could do more for actual sovereignty than another data centre. Data centres without Canadian data are just cold buildings.
- Move at the speed of the land grab. Frontier labs are not waiting for committee timelines. Every exclusive signed abroad is permanent. The window where Canada’s data is still Canada’s to direct is open now, and it closes deal by deal.
The seat is empty.
This is not an attack on the council. It is a strong group, and the strategy’s ambition is real. It is an observation about a blind spot: the thirteen people advising Canada’s AI future include no one whose full-time job is the thing models are made of.
Sovereignty is not a data centre. It is not a strategy document. It is the training data—who holds it, who clears it, who decides where it goes.
Someone should be at that table for the data. Right now, no one is.
The council membership and mandate referenced here come from the Prime Minister of Canada’s October 2, 2026 announcement ↗.
The C$2.4 billion and C$2 billion figures refer to the Government of Canada’s Budget 2024 AI announcements ↗.
Views expressed here are Archive Signal’s editorial position, not a claim that a dataset or licensing program currently exists.