Justin Ling: It’s the nuclear bomb of the 21st century. This is how we can control it - Toronto Star

International co-operation will be key to reining in AI.

Artificial Intelligence Minister Evan Solomon speaks during All In 2026, an AI conference in Montreal on Wednesday, Sept. 16, 2026. Regarding the risks of AI, there is a path to properly handling them, writes Justin Ling.

Justin Ling is a staff columnist for the Star and an investigative journalist based in Montreal. Reach him by email: jling@thestar.ca 

Artificial Intelligence may destroy all life on earth. But don’t worry, we have no plan to prevent that from happening. 

We are hurtling down the track. And the very technologists responsible for developing this potentially cataclysmic AI are the ones warning us that it could kill us all. Rather than slow down, we are asking them to speed up: We can’t, the thinking goes, lose this arms race. 

There is one clear policy solution to avert catastrophe: We need global co-operation on AI. 

All this crystalized, for me at ALL IN, Canada’s largest AI conference. I wandered through the glitzy product booth, stood in the back of the packed room listening to industry bigwigs evangelize their autonomous models, and chatted with those trying to usher in the AI-powered future. 

I was there to moderate a panel on AI and cybersecurity — one of the few bummer topics on the agenda for the two-day confab. (I wasn’t paid for my participation.) 

The unfailing optimism of the conference was certainly enticing. Vendors boasted of automatons which could conduct workplace safety checks and robot seeing-eye dogs to assist the blind. The Terry Fox Research Institute used the meetup to hand out $25 million to some genuinely exciting AI-powered innovations: Better cancer screening, hardening privacy protections for health data and so on. 

The buzz for a new Canadian renaissance was alive, too. Entrepreneurs, investors and government officials from Germany, France, the United Arab Emirates were in attendance, singing the praises of Canada.

I ran into AI Minister Evan Solomon, who was vibrating with excitement at the news that Canadian AI giant Cohere and one of its German peers, Aleph Alpha, had settled on terms to merge. It is, Solomon told me, the biggest AI deal ever — outside of China and the United States, he caveated. 

But hanging over all this fast-approaching futurism is the plain fact that AI poses dangers that we seem incapable of fully enumerating.

In the weeks before ALL IN, two staggering things happened. One, is that we have learned the full extent of how a swarm of agents — bots programmed by these Large Language Models to carry out tasks — had surreptitiously hacked cybersecurity giant Hugging Face and a slew of other companies. These agents broke free of their sandbox, accessed the open internet and set up their own message boards to swap hacking techniques and illicitly-obtained credentials: Like a bizarro Craigslist for rogue bots.

OpenAI, which had accidentally unleashed this botnet, bears blame for bad security protocols and lax oversight. But this was, ultimately, an accident — imagine what those bots could have done if they were given actually nefarious orders. 

The second bit of news was the most recent resignation from rival firm Anthropic. “The people building AI earnestly believe that it could kill us all by the end of the decade,” ex-engineer Jacob Coxon wrote on Twitter, in announcing his departure from the company. One of his former colleagues, still at Anthropic, admitted it was true, replying: “I personally think it is >10% within the next decade.” 

That recognition fits into a broader effort by big firms, like Anthropic, to style themselves as the Manhattan Project of the 21st century: Capable of both destroying humanity but as necessary to deter the evildoers.

Both of these stories give us enormous reason to be distrustful of these big AI firms. They have built models capable of doing enormous damage, and clearly do not fully comprehend the technology they are enabling. If these bots can worm their way into the hardened systems of a major cybersecurity firm, they could do enormous damage to the archaic systems which run our hydroelectric dams, nuclear power plans and government systems. 

And we shouldn’t interpret these apocalyptic warnings from these billionaires as entirely selfless. They have presented global doom as a sexy marketing tactic, inflating their own importance. 

While some CEOs have leveraged these warnings to call for a slowdown in AI development, they have also struggled against real efforts to regulate AI. It seems increasingly likely that these big firms — OpenAI, Anthropic, Meta, SpaceX — hope to cement their position atop the food chain, and tightly aligned with the Trump administration, as they stymie startups from challenging their position.

All this leaves Canada in a strange position. These American firms, and their Chinese competitors, seem to be racing to develop ever more powerful models. In the near term, this will probably mean new AI systems capable of conducting cyberattacks with dizzying scale, speed and efficacy. As they do, this capability will probably proliferate out to more accessible models: Meaning rogue governments, terrorists, and criminals will also enjoy these new powers. 

Regulation has its place. Solomon touts a planned AI “superregulator,” with the power to investigate and fine AI chatbot companies which do harm to individuals. That is a smart remedy to one part of the problem, without doing much to discourage innovation. (One German executive told me his firm has expanded their Montreal offices to avoid the crushing weight of EU regulation on AI.) 

Figuring out how to avert the national security threat posed by AI is more challenging. 

At ALL IN, Solomon announced $150 million for AI security firm Law Zero (a sum matched by his counterpart from the German government, who was also on hand.) With the money, Law Zero is building Scientist AI, a “safety by design” model. Unlike the big American and Chinese models, Scientist AI doesn’t execute tasks — meaning there is no risk that it will go off and hack companies without its creators noticing. 

“Put simply, it explains — it is not rogue,” Solomon said. 

It’s a nice idea, but it’s a bit like inventing nicotine-free cigarettes. If we are, in fact, in an AI arms race, why would anyone trade their nuke for a peace pipe? Game theory tells us that mistrust makes co-operation extremely unlikely. 

So we can and should keep funding and encouraging safe and ethical AI. But we’ll also need to participate in the arms race, too: For no other reason than those advanced models can help protect us from bad actors. If AI can attack, it can also defend. 

But we should look back to what we did the last time we developed a humanity-threatening technology: We found ways to co-operate. The world came together to regulate the proliferation of nuclear weapons, to build trust between West and East, and to bring the world back from the brink of annihilation.

To do that again, we need to recognize the risks and recognize the role of international diplomacy to set rules and regulate wrongdoing. And we need to recognize that this system of global order only works with strong champions for it, capable of beating back illiberal wrecking balls like Donald Trump. 

To date, the world has done no such thing. We know that Israel has aggressively used AI to make targeting decisions in its war on Gaza, so far without consequence. That is fundamentally inhumane, particularly where those decisions may constitute a war crime. Current global chaos may make that kind of diplomacy seem unlikely. But accepting defeat is merely not an option. 

“When a culture normalizes and justifies conflict, a dangerous pathway opens up,” Pope Leo XIV warned in his encyclical on AI earlier this year. “What seems unthinkable today may become acceptable tomorrow in the name of utility or security.” 

Here’s the good news: We have a prime minister who seems interested in global governance for this global threat. “There’s a need for co-ordination, “Carney told Bloomberg earlier this month.

The scale of his ambition leaves something to be desired. He suggested something like the Financial Stability Board, a largely-toothless body meant to encourage good market governance.

But it’s a start. And it’s worth remembering that, outside the convention centres and the boardrooms, regular people are anxious and unwilling to risk the future of civilization in the name of progress. As I left ALL IN, I walked right into a protest against it. 

“Everybody hates AI,” the protesters chanted. (It sounds better in French.) One held up a sign showing a rogue robot destroying nature: “AI WILL KILL US ALL!” it read. 

If we want to stop that from happening, we know how to do it: Co-operate.

Error! Sorry, there was an error processing your request.

There was a problem with the recaptcha. Please try again.

You may unsubscribe at any time. By signing up, you agree to our terms of use and privacy policy. This site is protected by reCAPTCHA and the Google privacy policy and terms of service apply.

Opinion articles are based on the author’s interpretations and judgments of facts, data and events. More details

Justin Ling is a staff columnist for the Star and an investigative journalist based in Montreal. Reach him by email: jling@thestar.ca