Share
In early June, Prime Minister Mark Carney stood in front of rows of medical workers to unveil his government’s new AI strategy. “AI for All” envisioned the rollout of artificial intelligence throughout Canadian society to make sectors more productive, provide benefits to Canadians, and ultimately increase Canada’s economic output.
For Carney, the choice to announce his strategy at a hospital seemed like a strategic one: he wanted the technology and his plan to increase adoption to be perceived positively by an increasingly skeptical public. In this way, AI would be associated with improvements to health care, better educational outcomes, and reduced demands for employees at work. It sought to paper over the growing list of social harms associated with AI—issues that have 68 per cent of Canadians calling for heavy regulation.
But alongside his optimistic advocacy for AI, Carney has been embarking on another major policy change with its own economic dimension: a military buildup the likes of which Canadians have not seen in decades. The government is committing up to $100 billion to buy a dozen new submarines, along with purchases of surveillance aircraft, icebreakers, joint-strike missiles, plans for new armoured vehicles, and plenty of additional military hardware. At some point it will make a long-delayed decision about new fighter jets too.
Many of these purchases are for traditional and very expensive military equipment, but as the wars in Ukraine and Iran, along with the genocide in Gaza, have shown, warfare is changing. Low-cost drones are an effective counter to traditional hardware and AI-infused weaponry can inflict devastating effects on their targets, which all too often include innocent civilians. These are technologies more often associated with the United States, Israel, Ukraine, or Iran. But Carney seems intent on ensuring Canada isn’t left behind on the more deadly and repressive markets for AI.
The deadly consequences of AI
For the past several years, Ukraine has been rapidly advancing its AI capabilities, equipping its Hornet drones with an AI-assisted targeting system that has contributed to its more frequent strikes on Russian infrastructure and supply lines. It has also used the technology for real-time planning and assessment of damage after strikes. These are uses that Western supporters of Ukraine would be likely to cheer because the technology is aiding an ally push back against an aggressor. But it can just as easily be deployed to depraved and deadly ends.
Take Gaza as an example. For the past several years, Israel has been utilizing AI targeting systems with names like Lavender and The Gospel to identify individuals and buildings with links to militants from Hamas or Palestinian Islamic Jihad. The effect has been to produce a list of targets much more rapidly than in previous military campaigns on the strip; targets that insiders say get “rubber stamped” for attack, with little checking of the reliability of the system’s outputs. In October 2023, a military colonel in charge of the target bank said those AI systems allowed the Israeli army to generate targets more quickly than it could attack them for the first time.
When those systems are combined with human decision-makers that disregard civilian life, the consequences are catastrophic. Journalists at +972 Magazine and Local Call documented how the Israeli army used AI systems identifying supposed militants in high-rise buildings to justify bringing down the entire structure, in a concerted campaign to make civilians feel the cost of the war. They also found that the Israeli army systematically attacked individuals in their homes, often at night when their families would be around them. In order to do this, they tracked suspected militants until they reached their family homes using yet another AI-powered software called “Where’s Daddy?”
The human cost of those actions are immense. Officially, Israel has killed more than 75,000 Palestinians in their attacks on Gaza, but experts estimate the real number is far higher, while 92 per cent of residential buildings have been flattened. But the carnage hasn’t ended in Gaza. As the United States and Israel turned their bombs toward Iran earlier this year, they once again employed AI systems to determine a wide range of targets across the country, hitting everything from nuclear facilities to a girls’ elementary school on the very first day of the war.
There’s little indication Canada is involved in these actions. But that doesn’t mean it won’t get to that point in the future. The Canadian government has continued sending weapons to Israel, despite the genocide it has carried out in Gaza, and to the United Arab Emirates, even as it fuels a genocide in Sudan and some of those weapons have been smuggled to Sudanese fighters. Canada claims to be a moral player on the world stage, but economics seems to trump everything else—and that’s especially the case in the Carney era.
Canada’s military-AI nexus
At the centre of the government’s strategy is a company called Cohere. In Carney’s announcement of the AI strategy, he touted how the company creates its own large-language models, making Canada one of only four countries with that capacity domestically. Yet Cohere is far from a model corporate citizen.
Cohere is being sued by 14 publishers, including the Toronto Star, for illegally using their copyrighted material to train its AI models and for allowing its products to generate fake articles attributed to those publications. The company’s attempt to dismiss the lawsuit was denied last year. Notably, Carney’s AI strategy has nothing to say about copyright infringement by AI companies.
Given the central position it holds in Carney’s plan to increase AI adoption, Cohere is poised to receive significant new public funding on top of the hundreds of millions it’s already received. Yet despite Carney’s talk of the importance of having Canadian companies build domestic data centres, Cohere chose a U.S. company called CoreWeave to develop its AI factory. Not only does CoreWeave have very questionable financial arrangements, but there were Canadian options available; they just would have taken a little longer to complete the project.
Just as U.S. AI companies like OpenAI, Anthropic, and the major players like Google and Meta have sought public contracts in recent years, so too has Cohere. It’s working with the Canadian and British governments to insert AI into the public sector—even as chatbots make up patients’ medical information, hallucinate immigration applications, and give Canadians incorrect tax advice. But it’s also been seeking those military contracts too.
Since December, it’s signed partnerships with the Canadian subsidiary of French defence company Thales, Swedish defence firm Saab, and German submarine marker ThyssenKrupp. Essentially, if the Canadian government is making a big new military procurement, Cohere seems to be in line for a lucrative AI contract as part of the promised domestic investment. It’s early days for Cohere’s move into military procurement, but as the government aims to ramp up defence spending, it wouldn’t be surprising for the company to see that sector as a growth area. It wouldn’t be the only one.
Earlier this year, researchers Mélina Poulin and Nicolas Chartier-Edwards warned about the militarization of the Canadian AI industry. Despite presenting itself as being concerned with principles of AI ethics, the industry was quietly shifting toward military and dual-use applications for their products, given where the business opportunities increasingly presented themselves. The experts expressed alarm at how that process was “taking place in a regulatory void, as Canada has no legislation on AI or autonomous weapons, and operates within an ethical vacuum that serves corporate opportunism and state complacency.”
A poor choice of partners
The government certainly isn’t providing much reassurance that the push to adopt AI won’t go sideways. In 2024, the Department of National Defence (DND) released a report chastising the department and the military for their slow and fragmented adoption of AI, setting out a strategy to streamline and accelerate that process. Under Carney, the pressure is only escalating, as he’s made it clear his focus is on adoption. Where his AI strategy seeks to address harms, it seems more designed around getting Canadians onside than really grappling with the extent of the technology’s drawbacks.
The DND has not shown itself to be a good judge of its partners. Between 2020 and 2025, it spent more than $45 million on a contract with Palantir, the U.S. data analytics firm co-founded by far-right billionaire Peter Thiel. Palantir CEO Alex Karp has defended the company’s work with U.S. Immigration and Customs Enforcement as it’s rounded up immigrants across the country, including dozens of Canadians, and called for U.S. tech companies to defend U.S. power, including by “making America more lethal.” Cohere had to disavow its previous work with Palantir earlier this year.
As the Canadian government sets out on a course to ramp up military spending and push AI at all costs, we need to be on guard for the ways those two industries are undoubtedly going to seek to collaborate regardless of the human cost of a greater application of AI in surveillance and warfare. We know the government is unlikely to stand in the way. Another pillar of the Carney agenda is a more conservative approach to crime, and there we can see AI is already being trialled in prisons and prompting concerns from legal experts.
This feels like a key part of the bait and switch underway. The government is trying to frame AI through a lens of social and commercial benefits that are mainly proving elusive, while trying to distract the public from the long list of harms that experts have identified and failures of implementation across the public and private sectors. But the harms can be so much greater, with people’s lives being shaped and even taken with the assistance of unreliable systems and authorities eager to use them to displace accountability.
Mark Carney is “all in” on a vision of AI that echoes industry narratives and seeks implementation at virtually any costs, including in the military. The public, however, is not behind him. And across various industries, the level of AI adoption has been slow, and the promised productivity gains have not materialized. One of the most significant exceptions to that rule, outside the tech sector itself, has been in the military-industrial complex. When we look back at the “AI revolution,” we may well see it primarily as a revolution in the technology governments use to kill, maim, and surveil.
About the author
Paris Marx
Paris Marx is a tech critic, author, podcaster, and international speaker. He hosts Tech Won’t Save Us and is the author of Hyperscale: The Ambition and Excess of Big Tech's Data Empires.
