Written by Nello Giuliani, a student volunteering with Mines Action Canada and graduating with a Bachelor of Arts in Computer Science from McGill University

The Standing Committee on Industry and Technology (INDU) held its last meeting on the role of AI in Canadian industry on June 16th, 2026, with a report expected once the House returns for the fall. The committee consulted experts and leaders in strategic Canadian industries to assess the benefits and risks of artificial intelligence in their respective fields, and to determine which legislative directions the House needs to take. Among the topics addressed were AI systems in the defence sector, autonomous weapons, and public trust in AI. After a brief overview of these topics, this article looks at how committee members and witnesses tackled them, and what needs to happen to prevent the use of autonomous weapons in warfare and to build trust in AI in the future.  

AI, similar to many technologies that have transformed industries on a global scale, is facing public scrutiny. As models become better at prediction and more architecturally complex, they often become less interpretable and more autonomous in their decision-making. These models can be weaponized in the form of autonomous weapons systems (AWS): weapons systems capable of searching for and engaging targets without human intervention. AWS raise ethical concerns as they can violate international humanitarian law, lower the cost of going to war, and complicate accountability among human actors. Many initiatives have called to preemptively ban the use of AWS and limit their development, including Stop Killer Robots, launched in 2013, which Mines Action Canada co-founded.

The topic of AWS was only addressed once during the study, when expert testimony was given by Université de Montréal professor Yoshua Bengio. When asked about balancing ethics and objectivity with competitiveness in military AI systems, he stated that the two are not at odds: there is demand for responsible AI. Even in a defence context, where a model's capabilities might grant battlefield or diplomatic leverage, ethics should be prioritized. To derisk AI systems, Bengio proposes a two-sided approach: scientific, where models are trained to respect legal redlines, and policy-oriented, including regulation for AWS, and international cooperation on AI uses in war. He also states that the first steps in AI safety lie in governmental and social change: “stronger, more democratic civil society and international institutions, to mitigate the potential commandeering and abuse of these technologies by governments or companies.”

Dr. Ali Dehghantanha, another expert witness consulted by the committee, reinforces these claims by arguing for human control over how frontier models are used. Dehghantanha, the Canada Research Chair in Cybersecurity and Threat Intelligence, blames the risk of AI-led national security breaches on the lack of control over AI systems post-deployment, and proposes that companies build a kind of checkpoint, or human control plane, that sits between an application and the AI model powering it, so nothing is fully autonomous. An adversary, potentially looking to conduct a long-range cyber-attack using an AI system, would be forced to go through said control layer. 

In terms of garnering public trust, the general theme among witnesses’ testimonies was the need for a clear national AI governance strategy. As stated by Teresa Scassa, the Canada Research Chair in Information Law and Policy, this could include a national coordinating body that would ensure transparency of regulatory activity and federal-provincial interoperability. There has not yet been an appropriate political response to the economic and social impact AI has had. A national plan can address citizens’ concerns with political safety nets that could reduce systemic social and economic effects. For example, if a sector of the economy sees significant downsizing, the state could subsidize new sectors to cover lost jobs; or if a specific knowledge system is marginalized by large language models, the state could increase public participation, including Indigenous voices, when developing AI policy. 

Other important factors that would help reduce public distrust involve making the technology more transparent, accessible, and fair. This includes modernizing data and privacy laws to enforce privacy rights and data sovereignty; elucidating privacy policies, training and validation data, and government actions seeking to ban or monitor accounts; and educational programs on AI to ease adoption. 

Overall, on the question of controlling weapons, the answer seems to be domestic regulation and international treaties to prevent misuse, and on the question of trust, transparent and actionable policies to safely integrate AI into the economy and society. These two answers are mutually reinforcing. However, crucial topics surrounding AWS and public trust are notably absent or only briefly addressed in the study and are of direct importance to AI risks and harms and potential regulatory measures.

First, a question that doesn’t yet have a clear answer: how much human control over autonomous weapons should be considered minimum? Researchers Roff and Moyes call this meaningful human control, and delineate it as a few concrete rules: systems have to be predictable and consistent enough for humans to trust; humans need to be confident in the information provided to them by the system; and someone has to be able to step in and stop the system in case of wrongdoing, with human accountability if something goes wrong. Current initiatives to integrate AI into military technology, such as the Department of National Defence’s 2026 contract with Larus, should make sure that human decision-making remains at the core of the technology.

It is also essential to look at the current uses of AI systems in ongoing wars to investigate how the same or similar systems are being developed or deployed domestically. The International Committee of the Red Cross states that IHL is not robust enough to address all the legal and ethical questions that AWS raise. This means international law needs to be bolstered through a legally binding international agreement on autonomous weapons, and at the national level through regulations that enforce ethical guidelines, such as import controls or third-party audits of AI systems.

Another important topic that deserves more attention is dual-use: models that can be used for both civilian and military purposes. The rise of general-purpose AI trained on large datasets makes it difficult to regulate models developed in the private sector because the intended use cases are benign, but the technology can still be deployed for malicious purposes. Thus, this might require developers of general-purpose AI systems to conduct dual-use risk audits before commercial release.

Finally, one direct cause of public mistrust in AI systems is simply a lack of transparency and a gap in communication coming from developers. Michael Geist’s proposal for an AI Transparency Act is one of the more tenable near-term interventions on public trust. The shortcomings of the Artificial Intelligence and Data Act (AIDA), an AI risk and transparency-focused bill that failed to pass in 2025, whose development restricted public participation, show that the process of drafting an AI Transparency Act would itself have to be transparent and inclusive. The AIDA did not do enough to protect Canadian workers and failed to uphold human rights, including those of First Nations, racialized communities, and gender minorities, all of whom should be equally represented in future legislation. 

Opportunities, Risks, and Regulation of AI in Canada’s Strategic Industries only scratches the surface of the risks and regulations, exposing the lack of focus on AI safety. Worryingly, one of the quasi-unaddressed topics is autonomous weapons. There needs to be more pressure to engage in the conversation on banning the use of AWS in war and limiting their development, and on making systems more trustworthy. We urge everyone to whom this cause holds importance to reach out to their Member of Parliament and bring the importance of building trust in AI and the urgency of halting killer robots to their attention.