Behind closed doors, AI policy writers, advisors, and advocates say it’s important that “middle powers,” potentially led by Canada, come together to shape guardrails for the technology.
Leaders from various government-associated AI policy organizations gathered at the British High Commission in Ottawa on Thursday to discuss the second-annual International AI Safety Report. Authored by over 100 AI experts and overseen by Canadian “AI godfather” Yoshua Bengio, the report found that risk management gaps could result in harm as AI capabilities continue to rapidly improve.
BetaKit attended the event under the Chatham House Rule, which means that commentary and information from the gathering can be shared publicly, provided that it is not attributed to specific people or organizations.
“If Mark Carney got up right now and said that we want to pause the development of AI capabilities, nothing would happen.”
In the intimate event’s conversations, speakers and attendees expressed that middle powers could create a new, influential authority over AI companies. Importantly, the authority would sit apart from the United States or China, two world superpowers many see as locked in an AI arms race with no interest in adding guardrails.
Bengio set the tone of the conversation in his opening remarks, which he made in a video call not held to the Chatham House Rule.
“In this race, countries like Canada have potentially a lot to gain, but also a lot to lose, especially if these advanced technologies end up being used unsafely, maliciously, or simply to dominate us economically,” Bengio said. “Safety is the only way that we can truly reap the benefits of AI, so Canada, along with its allies, can continue playing a key role in how we shape a safer trajectory in AI for the whole world.”
The event’s conversation had roots in Prime Minister Mark Carney’s speech in Davos earlier this year, which called for middle powers to work together against greater powers, though one policy advisor argued this speech has seen little meaningful follow-up. They added that middle powers can play an important role in pointing out the dilemmas that come with increasingly-powerful AI models and forming the safety mechanisms that could prevent “real lasting harm.”
“I think if Mark Carney got up right now and said that we want to pause the development of AI capabilities, nothing would happen,” the AI policy advisor said. “We don’t have the right chips right now to make that type of statement [and] I think that necessitates building a middle power coalition to make sure that we do.”
RELATED: Bengio-led AI safety report warns of growing risks amid rapid capability gains
Meanwhile, the United States can completely halt the release of a model, like it did with Anthropic last week, and the country might beat middle powers to setting the standard. CNBC reported on Wednesday that the CEOs of Anthropic and Google DeepMind called for a US-led coalition to protect against risks associated with AI at this week’s G7 Summit in France. Carney reportedly agreed that the US could lead an AI coalition.
However, another AI policy advocate speaking at the event said that, after a recent trip to California, they learned a number of people working on frontier AI models in companies like OpenAI, Anthropic, Google, Meta, and X.ai are “very concerned” about how quickly development is moving, and that they actually want “middle powers to step up to the plate.”
“They are concerned that the American government is not a reliable, good-faith actor,” the policy advocate said, stressing the Chatham House Rule. “They see that there is a need for regulation, both within the United States and in other countries, and they are there saying,‘we’re going to need middle powers to help us get to the right place.’”
Feature image courtesy Alex Riehl for BetaKit.