.png){kind=logo}
Canada’s federal government has opened a public consultation on how artificial intelligence systems should disclose their identity, limits, and activities. The process asks when Canadians should know they are dealing with AI. It also covers labels for AI-generated content and access to system information.
Artificial Intelligence Minister Evan Solomon announced the consultation through an online video created with AI tools. The video showed Toronto’s CN Tower launching into the sky like a rocket.
Solomon said, “AI can do extraordinary things.” He added that transparency matters as AI becomes part of everyday life. He also asked whether people can tell when they are speaking with machines. Another question covers whether users can identify AI-generated or altered material.
The consultation focuses on five areas. These include content labels, chatbot disclosures, system documents, incident reporting, and the tracking of AI agents. Canadians can submit their views until September 23, 2026. The government says feedback will guide policies, industry standards, voluntary codes, and possible legislation.
Officials also want to identify where existing rules already work. They are asking where further government action may be required.
The discussion paper asks how companies should report serious incidents linked to AI systems. It cites prompt injection attacks as one growing security concern. Such attacks can push an AI system to follow hidden or harmful instructions. The paper says they may cause data leaks, unauthorized actions, or financial fraud.
The government also wants clearer information about system abilities and limits. Officials are considering standard documents explaining intended uses, known risks, and safety controls. The paper points to problems caused by incorrect AI-generated legal citations. Canadian courts have reviewed cases involving invented legal authorities since 2024.
Officials are also asking whether developers should disclose details about training data. This could include the use of copyright-protected material. Technical options include digital watermarks and records showing where online content came from. However, the paper says users may remove or avoid these markers.
Ottawa is also considering whether incident reporting should become mandatory. The reporting process could cover security failures, financial losses, or unauthorized system actions.
The consultation gives special attention to AI agents. These systems can book meetings, send emails, complete forms, and make purchases. The government wants to know when agents should identify themselves to people and businesses. It also asks how users can track automated actions.
One example describes an agent that misreads a monthly shopping instruction. Instead of buying one package, it orders 12. The paper says responsibility may be unclear among the user, system operator, and original developer. Officials are examining how companies should record such transactions.
They are also studying how AI agents interact with other automated systems. Possible rules may cover permissions, human oversight, transaction records, and identity disclosure.
Meanwhile, Ottawa is advancing other measures linked to AI use. A digital safety bill would require platforms to label synthetic content that appears real.
A separate privacy bill would require organizations to explain automated decision-making in major personal matters. These decisions could include employment, credit, or access to services.
Another law passed in June criminalizes non-consensual sexual deepfakes. The government has also proposed rules addressing AI-generated election misinformation. The consultation supports Canada’s ‘AI for All’ strategy, launched on June 4, 2026. Protecting Canadians and democratic systems forms part of that plan.
The government plans to publish a ‘What We Heard’ report after the consultation closes. Public submissions may also appear online.
Also Read: Responsible AI: Can Transparency Keep Pace With Innovation?
