MONTREAL — Canada must band together with like-minded democracies to build and regulate AI, or face being dominated by the U.S., China and their tech giants, according to deep learning pioneer Yoshua Bengio.

The Université de Montréal professor has long called for governments to impose rules on AI developers to address the technology’s downsides, which he believes could range from more compelling online manipulation and disinformation to existential threats to humanity.

Regulations should require that firms be transparent about the capabilities of their AI systems, the data on which they are trained, the resources they consume, the risks they pose and the processes in place to address those issues, Bengio told The Logic on stage at the Attention Forum, a conference on digital governance in Montreal. 

That requires collective international action, according to Bengio. “Canada is too small by itself to have an influence,” he said. “It should be very busy forming partnerships with other countries.” In addition to rule-making, he called for Canada, the European Union and other allies to work together to develop safer and more ethical AI systems.

While the U.S., China and their tech firms currently lead the field, Bengio insisted middle powers can still work together to compete at the frontier. It would be a “huge mistake” to conclude that it’s too late to catch up, he said. 

Bengio has spearheaded one international effort to align nations on AI, to mixed success. In November 2023, the British government named the computer scientist as lead author of a “state-of-the-science” report on the capabilities of the most advanced AI systems. The study was intended to inform policymaking in dozens of countries—including Canada, the U.S. and China—that had committed to co-operate more closely on AI risks and safety. 

Bengio delivered his report in January, warning that the technology’s rapid improvement was increasing its dangers. It was met with a muted response as nations have shifted to jostling for AI dominance.

U.S. President Donald Trump has said the U.S. wants “unquestioned and unchallenged global technological dominance,” and his administration has warned other countries against regulating its tech giants. China also aims to be the world’s leading AI power. Meanwhile, France, Germany and other European states have sought to dial back the EU AI Act over concerns it would constrain their own tech sectors.

Bengio blamed the shift on tech giants lobbying policymakers to avoid regulation. He cited the U.S. political action committee Leading the Future, launched in August 2025 with over US$100 million from Silicon Valley donors to support “pro-AI” candidates and counter “anti-innovation narratives,” as one example of how the political mood had been changed. 

Policymakers and the public should not treat AI as just another product, technology or economic force, because it could quickly become much more significant than that, Bengio said. “It’s a game of power where the people or the countries or the corporations that control [it] have the potential to dominate the world,” he said, predicting that Canada could see its democracy weakened and its homegrown firms wiped out if it can’t compete on AI. 

On stage at the Attention Forum shortly after Bengio, AI Minister Evan Solomon reiterated that Ottawa aims to chart a middle course on regulating the technology. The U.S. and China see “any form of regulation [as] a constraint on their national security and their industry,” he said, while the EU’s regulatory model has been criticized as “too stifling on innovation.” 

Solomon once again said he intends to table privacy legislation, but not reintroduce the previous government’s proposed AI law. “We need to empower our businesses so they can remain competitive with this tool,” he said.

Bengio is now co-president and scientific director of LawZero, a new non-profit he founded to develop an AI tool for research that can also act as a guardrail for other systems. While governments must set new rules, regulators don’t have the necessary expertise to evaluate and police the most advanced models, he said. 

To fill the gap, Bengio recommended AI developers be required to have their products assessed by independent auditors or insurers to identify risks and whether they’re doing enough to mitigate them. “It would make the dollars of those companies work for the public good,” he said.