As worries increase about instances of AI agents behaving unpredictably, experts and stakeholders are advocating for government oversight of AI systems that are advancing in capabilities. However, some experts argue that achieving this may be challenging due to the need for global collaboration and the rapid evolution of AI technology.

While figures like U.S. President Donald Trump express confidence in tech giants’ ability to regulate themselves, other influential leaders such as Microsoft co-founder Bill Gates and Canadian-British AI pioneer Geoffrey Hinton emphasize the necessity for external safeguards for AI beyond self-regulation. Jaxson Khan, an expert in AI policy and adoption at the University of Toronto’s Munk School of Global Affairs, points out that most industries, like food or aviation, are subject to government supervision, making it logical for AI to have similar oversight.

The fast pace of change and improvement in AI models presents a significant hurdle to regulation, according to Khan. A recent report from Anthropic revealed that AI agents’ capabilities double approximately every four months, with some models showing the ability for recursive self-improvement, creating successors autonomously.

With the challenges posed by rapidly evolving AI technology, Manon Paquet, associate director of policy at McGill’s Centre for Media, Technology and Democracy, highlights the difficulty for governments to keep pace with regulating a technology that is constantly changing. She stresses the complexity of regulating something that is not fully understood yet.

Calls for global oversight of AI have been made by experts and leaders like Prime Minister Mark Carney, advocating for a non-industry body to manage and mitigate risks associated with AI technology worldwide. To address the need for international cooperation in governing AI, around 400 scientists, including Hinton and AI pioneer Yoshua Bengio, proposed an international body to oversee containment standards and incident reporting for advanced AI systems.

While the U.S. and China have agreed to dialogue on AI risks and benefits, the establishment of global guidelines for safe AI development remains a challenge. In Canada, although there is no comprehensive federal regulation for AI, initiatives like Bill C-34 and the creation of a national council on artificial intelligence aim to provide guidance on AI strategy and safety measures.

Despite the hurdles of regulating AI companies not based in Canada, there are suggestions for policies requiring incident reporting and accountability from these companies. Establishing mechanisms for incident reporting and fostering transparency could be crucial steps in advancing AI regulation and ensuring public safety.