Safety institute announces research projects as global focus shifts to AI adoption

OTTAWA — The Canadian Artificial Intelligence Safety Institute will fund research projects focusing on misinformation, generative AI and the safety of autonomous systems.

It said the initial 10 projects will receive $100,000 each from its research program.

They include a project led by AI pioneer Yoshua Bengio that will look at the hidden reasons driving decisions by large language models, a type of generative AI focused on text.

Other projects receiving federal funding include efforts to develop a tool to detect foreign interference, research on how to reduce the risk of models memorizing private information obtained from training datasets, and research on ways to align AI systems with human values.

Other initiatives will focus on AI-assisted coding and “agentic” AI, or AI that acts autonomously with limited human intervention. One project will involve autonomous chemistry laboratories.

Catherine Régis, co-director of the institute’s research program, said AI safety is a new field.

“There’s a lot of things that we still don’t know. That’s why the research program is so important,” she said.

“We are kind of pushing the envelope of what is AI safety, how the Canadian research community can really contribute in defining this, and making sure that we are two steps ahead … in preventing what could happen in this sphere that is moving so fast right now.”

The Canadian AI Safety Institute was launched last year. It’s part of a global network of publicly backed safety institutes that emerged following a wave of calls for AI regulation from experts — some of whom have warned that the technology could pose an existential threat to humanity.

But there has been a global shift among governments toward a focus on AI adoption over safety initiatives in recent months.

Régis said that shift does worry her but Canada can focus on its strength in research as a form of soft power.

Canada’s widely recognized research strength in AI comes in part from three national research institutes and the so-called “godfathers” of AI — Bengio and Geoffrey Hinton.

The previous Liberal government introduced an AI regulation bill targeting “high-impact” systems, but the bill did not become law before the election was called.

Testifying at a Parliamentary committee last year, Bengio urged the government to put a regulation law in place immediately, even if it’s not perfect.

He said the bill should cover technology that poses national security and societal threats — such as systems that bad actors could use to design dangerous cyberattacks and weapons, or systems that find ways to self-replicate despite programming instructions to the contrary.

The current Liberal government has not indicated whether it will bring that legislation back or bring forward a similar bill.

It has said it plans to prioritize AI’s economic potential as it hosts the upcoming G7 summit. In May, it created an AI ministry and named former journalist and rookie MP Evan Solomon to the post.

The press release about the grants included a statement from Solomon in which he said the funds “support world-class Canadian researchers who are tackling some of the most pressing challenges in AI, including safety and ethics, accountability and real-world impact.”

He said the government is “not only reinforcing our leadership in responsible AI but also building a more productive, people-focused economy and helping Canada lead the G7 in growth, trust and innovation.”

But it’s not clear where Solomon stands on the question of how to balance AI safety with promoting the technology as industrial policy. A month after being appointed, Solomon has declined to answer reporters’ questions about the nature of his job.

This report by The Canadian Press was first published June 9, 2025.

Anja Karadeglija, The Canadian Press