Good morning! Welcome to the Canadian AI Newsletter, a weekly rundown for founders, operators and investors.
Was this forwarded to you? Subscribe below.
I am Raif Barbaros, Partner at Mistral Venture Partners. Views are my own.
I also run the Canadian AI Index, a free database of Canadian AI companies, founders, rounds and investors.
Sequoia led a US$60M Series B in Mecka, which builds robot training data with a team mostly based in Toronto. Polygrade came out of stealth with US$6.3M, Zeta Global is buying Toronto’s Senso, and Cohere is opening an office in Ottawa. OSFI says it may write AI rules for banks, and AI data-centre proposals are drawing pushback in Edmonton and New Brunswick. Also in this issue: young people turning to AI for mental-health support, and AI reshaping entry-level cyber jobs. Let’s get into it.
💰 Deals & Milestones
Mecka (Toronto and New York) raised a US$60M Series B led by Sequoia on October 7.
Nvidia, Samsung and the venture arms of Microsoft and Qualcomm joined. Mecka records people doing everyday tasks to build training data for robots, and says it passed a US$100M revenue run-rate in June.
Headquarters is New York, but most of the team is in Toronto.
Polygrade (Toronto) came out of stealth with US$6.3M in seed funding on October 6 for AI agents that process home-warranty claims.
Construct Capital put in US$3M in July, after US$3.3M from American Family Ventures, Nine Four, Hustle Fund, Room and Pillar in early 2025. CEO David Steckel founded Setter, which Thumbtack bought in 2020.
Senso (Toronto) agreed to be acquired by Zeta Global, announced October 5; I missed it last week.
Senso’s tech already powered Zeta’s product for tracking how brands show up in ChatGPT, Gemini and Claude answers. Terms weren’t disclosed.
AirMatrix (Ottawa) got its largest NRC IRAP project to date on October 9, through IRAP’s defence program, to fuse radar, radio-frequency and camera data with AI for low-altitude airspace security.
The amount wasn’t disclosed.
CyberWorks Robotics (Markham) will supply autonomous vehicles to the Canadian Armed Forces under a contract worth more than $1M.
The amended DND contract was awarded in late August and runs to February 2027, with a path to up to $7M more.
SenseHuman AI (Edmonton) is building a hardhat with AI cameras that watch a worker’s blind spots.
CEO Adil Muhammad founded it last year after AI research at U of A. It has customers and is raising to scale manufacturing.
CoLab (St. John’s) signed a global contract with Vestas on October 8 for AI-powered design review across Vestas R&D and its suppliers.
Vestas is also testing AutoReview, CoLab’s AI checker for design-for-manufacturing feedback.
BioRender (Toronto) launched Leo on October 6, an AI agent that builds scientific figures from a library of vetted icons.
🏢 Large Companies
Cohere is opening an office in Ottawa, its third in Canada after Toronto and Montréal. (Disclosure: Cohere is a Mistral Venture Partners portfolio company.)
Nearly 30 Ottawa-area staff in sales, engineering, infrastructure and security will use the space near Lansdowne Park.
EDC wants to package and finance stacks of Canadian tech, AI compute included, for foreign buyers, CEO Alison Nankivell told The Logic on October 8.
EDC is working with Global Affairs’ new Strategic Exports Office. There’s no program or dollar figure yet.
🔬 Research
CIFAR named seven Canada CIFAR AI Chairs on October 6, the first cohort under the AI for All strategy, backed by nearly $4.5M.
Six are new and one is renewed. The researchers are at SFU, UBC, Concordia, McGill and HEC Montréal, and work through Amii and Mila.
UBC and the Vector Institute tested whether AI can spot future depression in essays children wrote at age 11.
Main finding: in a British study that followed the same people from childhood, a simple model using six basic facts about each child beat every AI model at predicting depression symptoms at age 23 (0.737 vs 0.670 on the standard accuracy score). Adding the AI’s reading of the essays didn’t improve it.
So what: health-AI companies that predict outcomes from text should show what the text adds over a simple baseline.
The Vector Institute had 20 AI models play the full video game Slay the Spire at its hardest setting, where early choices decide whether you win much later.
Main finding: the best models win the first stage as often as expert players, then fall behind. All of them play too safe. They hoard health instead of spending it to get stronger, and the notes they leave themselves between games are often wrong.
So what: long-range planning is still a weak spot for AI agents.
🏛️ Policy
OSFI is considering an AI “safety code” with minimum standards for banks, Superintendent Peter Routledge said on October 7.
He pointed to Mythos and other frontier models that could be used for cyberattacks, and to AI agents working with little human oversight. There’s no draft yet, and a consultation could come in spring.
The next day, OSFI’s fall risk outlook focused on how frontier AI amplifies cyber, third-party and reputational risk.
🖥️ Data Centres
Walton Global (U.S.) asked Edmonton to amend the plan for its Energy and Technology Park to allow a hyperscale-ready AI data centre in the city’s far northeast, CBC reported on October 8.
The site borders Sturgeon County, where Meta plans its $13B data centre.
FTL (Fredericton) pitched Miramichi council on a $2.7B, 100 MW AI-capable data centre that it says would create 600 permanent jobs.
A day earlier, Green deputy leader Megan Mitton asked the province to disclose how many AI data centres are proposed and where, after documents showed talks on a second large data centre in Saint John.
📊 Data
Jack.org, the Canadian youth mental-health charity, found one in three young people use generative AI at least monthly for mental-health support, and nearly half haven’t told anyone.
The survey covered 395 young people in Jack.org’s network. 78% reported a barrier to care such as cost, wait times or stigma.
TMU’s Rogers Cybersecure Catalyst found 93% of employers say human skills matter more as AI takes over entry-level cyber work such as alert triage.
The Canadian AI Index counts 36 robotics and physical AI companies. Twelve are in B.C., more than double Toronto’s five.
📰 In brief
Mila and Inria signed a four-year research agreement.
Genome Alberta and the province put $6M into seven projects that use genomics and AI on livestock, crop disease and pollinators.
U of A researchers built SABER, an AI robot that speeds up seed inspection for Edmonton’s 20/20 Seed Labs.
Genesis (St. John’s) launched T-Minus30, a free 30-day sprint for anyone in Newfoundland and Labrador who wants to build with AI tools.
Nova Scotia doctors told CBC that patients now check ChatGPT Health and Claude before their appointments, often without saying so.
Adastra (Toronto) bought Infinite Services, a Polish data and AI consultancy.
OpsGuru was named an OpenAI Select Partner for running OpenAI models on AWS for regulated industries.
New on the Canadian AI Index: Polygrade, AirMatrix, CyberWorks Robotics, SenseHuman AI, BioRender and Tidal Point (Ottawa), which makes audit-grade AI for enterprise risk management.
FedDev Ontario will announce AI funding in London, Ont. today at 1:30 p.m.
Thank you for reading! Subscribe, and please share any feedback 🙏🏽
— Raif