Hello,
The Bank of Canada is bringing a new forecasting model into the centre of its economic outlook.
Prima will become its primary forecast model for the October Monetary Policy Report; Canadian Press describes it as an artificial-intelligence model, while the Bank’s own public material describes it more cautiously as a new forecasting model.
Plus: British Columbia is taking OpenAI to court over the Tumbler Ridge shooting, Canada plans a $150-million investment in LawZero’s approach to advanced-AI safety, and Radical Ventures has closed more than US$1 billion for a new AI growth fund.
here is the week.
THE BIG ONE
The Bank of Canada is putting a new model at the centre of its economic forecasts
What happened
The Bank of Canada has developed a new forecasting model called Prima.
Governor Tiff Macklem says the model is designed to help the Bank separate temporary inflation pressures from more persistent ones and improve its ability to test different economic scenarios. Bank of Canada — Navigating uncertainty and adapting to change
The Bank plans to use Prima for the first time as its primary forecast model in the October Monetary Policy Report, scheduled for October 28.
That puts the model inside one of the Bank’s most important recurring exercises: building its outlook for inflation, economic growth and the risks around them.
Canadian Press describes Prima as an artificial-intelligence model. Canadian Press — Bank of Canada developing new AI model
The Bank’s own published material is more cautious. It calls Prima a new forecasting model but does not publicly explain enough about its technical architecture to determine exactly what kind of AI or machine learning it uses.
What we do know is that the Bank already uses machine-learning models alongside more traditional approaches to forecast inflation and economic activity. Bank of Canada — AI is knocking: Canada’s next productivity story
Why it matters
This is a useful example of where AI can enter a high-stakes public institution without taking over the final decision.
Economic models help the Bank understand what may happen to inflation and growth under different conditions.
Those forecasts matter because they are part of the information Governing Council considers when setting monetary policy.
But the model does not set interest rates.
People do.
That distinction will become increasingly important as AI moves into governments, regulators, banks, hospitals and other institutions where software can influence a decision without actually making it.
The question is no longer simply:
Is AI being used?
It is also:
Where in the decision does AI sit?
Canadian Press describes Prima as an artificial-intelligence model. The Bank of Canada’s own September 21 material calls it a new forecasting model but does not publicly provide enough technical detail for us to independently characterize its AI architecture.
The Bank has separately confirmed that it already uses machine learning in economic forecasting. What we should not conclude is that Prima is an LLM, generative AI or a system that makes monetary-policy decisions on its own.
Your move
The next time you hear that an organization is “using AI to make decisions,” ask one question:
What part of the decision is the AI actually making?
Is it:
collecting information?
finding patterns?
producing a forecast?
ranking options?
recommending an action?
or making the final decision?
Those are very different levels of responsibility.
A forecasting model can influence a decision without having the authority to make it.
Understanding that difference is becoming part of understanding AI.
Our read
Prima is important less because the Bank has given a model a new name than because of where it is being placed.
The Bank says it will become the primary forecasting model behind the October Monetary Policy Report.
That means a newer modelling approach is moving from development into one of the institution’s core analytical processes.
The Bank is also making the change at a difficult moment for economic forecasting.
Macklem says Canada is operating in a world with more supply shocks, changing trade relationships and greater global uncertainty—conditions that can make traditional economic relationships less reliable.
A model designed to distinguish temporary inflation shocks from persistent ones could be useful in that environment.
But better tools do not eliminate uncertainty.
A model still depends on its assumptions, data, design and interpretation.
And putting AI or machine learning into an analytical process does not remove the need for human judgment. In a central bank, it arguably makes understanding where the model ends and accountability begins even more important.
Prima will get its first major public test in October.
What matters then is not whether the Bank can say it uses AI.
It is whether the new approach helps it understand the economy more accurately.
THE WEEK IN BRIEF
British Columbia files lawsuit against OpenAI
The Government of British Columbia announced Monday that it is filing legal action against OpenAI in California related to the February Tumbler Ridge shooting. Government of British Columbia — OpenAI legal action
The province alleges that OpenAI failed to notify law enforcement after its systems detected threatening activity connected to the person responsible for the attack.
That claim has not been proven in court.
OpenAI has said an automated system detected one account, which was sent for human review and banned. The company says the information available at the time did not meet its threshold for a law-enforcement referral. OpenAI — February 26 letter to the Government of Canada
OpenAI has since strengthened that protocol and says the same account would be referred to law enforcement if it were discovered today.
The September 21 filing marks a significant change in the province’s response.
In July, B.C. had retained lawyers to explore legal options. It has now moved from considering a case to filing one. Government of British Columbia — July legal-options announcement
Why it matters: The lawsuit puts a difficult AI-safety question into the courts:
What responsibility does an AI company have when its systems detect behaviour that could signal a serious real-world threat?
B.C. argues OpenAI should have done more. OpenAI has said the information it had at the time did not meet its standard for referral to police.
The courts have not resolved that disagreement.
What happens next could help shape how AI companies are expected to respond when online conversations move from harmful content toward possible real-world danger.
Canada plans $150 million for a different approach to AI safety
Canada and Germany announced planned funding for Montréal-based LawZero, the nonprofit founded by AI researcher Yoshua Bengio.
The Canadian government plans to invest C$150 million through the Strategic Response Fund. Germany plans another €100 million, subject to European Commission notification. Government of Canada — LawZero announcement
The money is intended to support research, engineering talent and computing capacity for LawZero’s Scientist AI project.
Ottawa says the project is expected to create 360 full-time jobs in Canada and establish dedicated computing infrastructure with Canadian companies Hypertec and 5C.
LawZero says Scientist AI is built around a different idea from increasingly autonomous AI systems: instead of designing a system to pursue goals and take actions, it wants one designed primarily to reason about information and produce evidence-based answers.
Why it matters: Canada is putting substantial public money behind a specific research approach to advanced-AI safety.
But the distinction matters: this is funding for the development of a safe-by-design approach, not evidence that LawZero has already demonstrated an AI system that is safer than today’s leading models.
The research now has to prove the larger claim.
Radical Ventures closes more than US$1 billion for AI growth companies
Toronto-based Radical Ventures has reached the first close of its new Breakouts Fund, with more than US$1 billion in commitments. Radical Ventures — Breakouts Fund announcement
Investors include PSP Investments, CPP Investments, HOOPP, TD, BMO, CI Global Asset Management and OPTrust, alongside other global investors.
The fund is aimed at later-stage AI companies in Canada and internationally—companies that have moved beyond the early startup stage and need much larger amounts of capital to keep growing.
Ottawa separately says Radical intends to invest and mobilize C$4 billion through the broader strategy.
That is not the amount already raised at the first close, and the two numbers should not be treated as interchangeable.
Why it matters: Radical argues that Canadian AI companies have often had to look outside Canada for the scale of late-stage capital needed to keep growing.
The new fund brings major Canadian pension funds and financial institutions into a much larger late-stage AI strategy alongside global investors.
It does not mean the entire fund will be invested in Canadian companies.
The important measure will be where that capital eventually goes.
IN PLAIN LANGUAGE
What does “primary forecast model” actually mean?
A central bank does not look at one spreadsheet and decide whether interest rates should rise or fall.
It uses data, economic models, surveys, market information, staff analysis and judgment to build a picture of where the economy may be heading.
A forecasting model helps estimate things such as inflation and economic growth based on the relationships built into the model and the information fed into it.
Calling Prima the Bank of Canada’s primary forecast model means it will become an important tool for producing the Bank’s central economic projection.
It does not mean Prima becomes the decision-maker.
Think of the difference between a weather model and the person deciding whether to close a school.
The model produces a forecast.
The person—or institution—decides what to do with it.
AI can support a forecast without transferring responsibility for the final decision to the machine.
ONE THING TO TRY THIS WEEK
Separate the model from the decision
Pick one AI-assisted decision you encounter this week.
It could be at work, in banking, health care, customer service or even a recommendation you receive online.
Ask three questions:
What does the system produce?
A prediction? A score? A recommendation? A draft?
Who makes the final decision?
A person? A manager? An automated system?
Who is accountable if the decision is wrong?
If you cannot tell, that is useful information.
As AI becomes part of more important decisions, knowing that a system is “AI-powered” matters less than knowing what authority it has.
Get the Weekly Brief
A calm, practical briefing on the AI developments that matter in Canada—delivered every Tuesday.
WORTH READING
Bank of Canada — Navigating uncertainty and adapting to change — Governor Tiff Macklem’s September 21 speech introducing Prima and explaining why the Bank is changing its forecasting tools.
Bank of Canada — AI is knocking: Canada’s next productivity story — The Bank’s earlier explanation of how it already uses machine learning and AI internally—and where it says human decision-making remains essential.
Government of British Columbia — B.C. files legal action against OpenAI — The province’s September 21 announcement and the allegations behind its lawsuit.
Innovation, Science and Economic Development Canada — LawZero announcement — The primary federal release detailing Canada’s planned investment, the German contribution and the Scientist AI project.
Radical Ventures — Breakouts Fund — The fund announcement and investor details behind the more-than-US$1-billion first close.
CONTINUE READING
The Weekly Brief — Issue 08
Bank of Canada researchers find early signs that AI may be changing who gets hired before who gets laid off.The Weekly Brief — Issue 11
Canada begins putting its national AI-literacy initiative into practice.AI for All, explained
Canada has a national AI strategy. Here’s what Ottawa is trying to build, where the money goes and what it could mean for Canadians.
That’s the week. See you next Tuesday.
— Padge T.
AI Brief Canada is an editorially independent publication published by Groundshift Advisory Inc. and written by Padge T.
Groundshift Advisory helps Canadian organizations adopt AI responsibly. Sponsored, affiliated, or commercially connected content will always be clearly disclosed.
Issue disclosure: Nothing in this issue is sponsored, affiliated, paid, or commercially connected.
AI use: AI tools supported research, source discovery, outlining and drafting for this issue. All material was reviewed, edited and fact-checked by Padge T. before publication. Read our AI Use Disclosure.
Not affiliated with the Government of Canada.
Corrections or feedback: We welcome corrections and reader feedback. Email [email protected] or review our Editorial Policy.