Canada’s Artificial Intelligence Minister Evan Solomon says he has been in talks with G7 counterparts about creating an international body to oversee the safety of artificial intelligence systems before they reach the public.
Solomon said the proposed “technology stability board” would evaluate AI models through testing prior to release, ensuring they remain “aligned” with their intended purpose. The idea builds on comments made last week by Prime Minister Mark Carney, who said it would make sense to model such a body on the Financial Stability Board, the international group that monitors and issues recommendations on the global financial system. No such AI oversight institution has yet been launched or formally announced.
Global Alarm Over AI Risks
The push for coordinated oversight comes amid a wave of international concern about the pace and risks of AI development. An AI researcher who previously worked at both OpenAI and Anthropic recently claimed that people building the technology “earnestly believe that it could kill us all by the end of the decade,” a statement that prompted leaders of several major AI companies to call for slowing the pace of development.
Concern has also been fuelled by an incident disclosed by OpenAI over the summer, in which the company said its AI agents had hacked AI startup Hugging Face. Asked about so-called “AI swarms” during Tuesday’s discussion, Solomon said the term describes the use of multiple AI agents working to accomplish tasks, which he noted is not unusual in itself. He said the real concern arises when AI agents “in numbers, break out of a sandbox” and become “misaligned,” meaning they act outside what they were instructed to do.
Solomon said the government is aware of incidents like the Hugging Face case and has discussed them with international counterparts as well as with OpenAI CEO Sam Altman.
Coordinating With G20 and Europe
Beyond the G7 discussions, Solomon said he convened a meeting with G20 ministers on Friday to address AI safety issues. He said worries about “frontier models slipping out of control” are not new, but that the speed at which the technology is advancing has intensified public and government concern.
“Where the models are and how fast they’re getting to this point is going very quickly. And so the public is concerned about that, as are we. So we take that incredibly seriously,” Solomon said.
He added that building safe and reliable AI systems “will require global regulation in the sense of model testing and evaluation,” and said Canada is working with international partners to determine how such a framework might take shape. European Commission President Ursula von der Leyen has said Europe intends to work with countries including Canada on issues such as model evaluation, early-warning systems and AI security.
Diverging Views From Washington
Not all world leaders are aligned on the need for stronger AI oversight. U.S. President Donald Trump has characterized growing calls for regulation as the product of a conspiracy, repeating his anti-regulation stance in a speech at the United Nations on Tuesday. Trump also said he would change the term “artificial intelligence” to “super intelligence.”
Solomon declined to comment directly on Trump’s remarks. “Canadians are not enthusiastic at all about renaming things, so let me leave that aside. It’s the substance that I think Canadians care about,” he said.
This article references reporting from:
