On Tuesday Canada's minister of artificial intelligence, Evan Solomon, announced an extra C$500 million over five years for the Regional Artificial Intelligence Initiative. That takes the program to C$700 million. The money runs through the country's regional development agencies and goes to businesses that want to put AI to work, and to Canadian AI firms trying to sell and scale what they build.

Solomon was blunt about the goal. "The question is not whether AI will reshape our economy. It is whether Canada will capture the opportunity," he said, so that the technology, talent and value "remain anchored in Canada and under Canadian control." Health and life sciences, energy, farming, manufacturing and robotics are first in line. (Government of Canada release; Fathom)

The dates matter if you operate in Canada. In southern Ontario, applications are open until January 19, 2027, but projects with costs before March 31 must be in by November 24. Since the program started in 2024 it has backed 218 projects, which Ottawa says created or kept more than 1,100 jobs and drew about C$1.77 of outside money for every dollar it put in.

The cheques go to companies, with a special nod to small and mid-sized firms, and it sits inside AI for All, the national strategy launched in June. Ministers from Quebec to the Prairies to British Columbia lined up with quotes. When every region claims a slice, the program tends to stick around. (release)

Solomon spent the same Tuesday in London. Canada and the United Kingdom renewed their AI security partnership, first signed in 2024. The Canadian AI Safety Institute and the British AI Security Institute will share findings on frontier risks, compare how they test models, explore staff swaps and joint evaluations, and work on common technical standards.

One line stands out. The two governments will widen information sharing "including through appropriate Five Eyes channels for sensitive reporting." That treats serious AI risk the way allies treat intelligence. Put the two announcements side by side and you get Canada's position: spend to adopt, and keep watching what the models can do. (Canada-UK statement; GOV.UK)

Others are building the watching part now. The United Arab Emirates says every AI agent in its federal government will have to plug into a central "control tower" that logs its actions and decisions, audited by people and by software. Its AI minister, Omar Al Olama, made the point while the country pushes to run half of federal services on AI agents within two years.

Singapore went the testing route on Wednesday. Its media and tech regulator, IMDA, released a starter kit for checking AI agents before they go live: what tools they use, what data they touch, whether they stay inside their permissions. It also named the first five accredited third-party AI testers, PwC Singapore among them. Adoption money and oversight are arriving together. (Gulf News; The Straits Times; AsiaOne)

Yesterday's The Real-Time Ask stays sent: Singapore asked the labs to report incidents as they happen. Tuesday's The Near Miss and Monday's The Brake stay sent too.

So do The Panel, The Decree, The Ten and The Guidelines. Read today against the Real-Time Ask. Last week was about who hears when AI goes wrong. This week a government is paying to put more of it to work.

We have seen this movie with cloud. Governments and vendors paid companies to move fast, and the security reviews and exit plans came years later, usually after something broke. Grants speed up the buying decision. They rarely pay for the logging, the testing or the person who owns the risk.

The difference this time is that the tools act. An AI agent that books, pays or writes code on your behalf needs the same controls the UAE is wiring in and Singapore is testing for. If a subsidy gets the tool through the door faster than the controls, the board has bought the risk at a discount. (Gulf News)

Put four questions to management. Are we applying for any public AI money, in Canada or elsewhere, and what does it commit us to? Who approves an AI tool before it goes live, and does a grant-funded one skip any step? Which of our AI agents can act without a person checking? And could we show a regulator a log of what they did?

Then check the calendar. If you have Canadian operations, the November 24 deadline is close enough to rush a project. Rushed projects are where controls get left for later. Ask for the control plan to arrive with the application, not after the money does. (release)

The Move

Governments are now paying companies to adopt AI and, in the same week, building the logs and tests to watch it. Boards should do both at once.

Have the chief financial officer list every AI grant or subsidy the company is chasing this year, and have the chief information security officer attach a one-page control plan to each: who owns it, what it can touch, and how its actions are logged.

No control plan, no application.

Seven bullets. One move.