The European Union just flipped the switch on mandatory AI transparency, and the clock for Canada to follow suit now ticks louder than ever. As of Sunday, any artificial intelligence system operating within the bloc must unmistakably signal to users that they are engaging with a machine, not a person.
These long-awaited obligations form part of the EU’s sweeping AI Act, originally passed in August 2024 but granted a two-year runway before enforcement began. The rules split responsibilities between providers and deployers. Companies that build AI models, such as OpenAI, must engineer systems that alert users upon interaction. Meanwhile, platforms using those models need to tag deepfake videos, audio, and images with clear labels, particularly when the content involves real individuals or matters of public concern.
Machine-readable markers embedded in AI-generated or manipulated media represent a cornerstone of the framework. The European Commission published an optional set of standardized icons to simplify compliance. A fully synthetic video would carry an “AI Generated” badge. An authentic photo altered to swap a politician’s face would display “AI Modified.” Platforms like TikTok and Instagram already deploy their own labeling systems, which the EU deems acceptable provided they meet the law’s intent.
Penalties give the rules serious teeth. Violators face fines reaching 15 million euros, roughly 24.2 million Canadian dollars, or three percent of global annual turnover. Even EU institutions themselves risk penalties up to 750,000 euros. Notably, satirical and artistic works receive carve-outs from labeling mandates, a deliberate attempt to avoid drowning users in meaningless notifications.
Canada now scrambles to close the gap. Ottawa launched a public consultation last month, accepting submissions until September 23. The exercise probes five transparency priorities: detecting synthetic content, notifying individuals when they encounter AI, explaining system capabilities and limits, tracking serious incidents, and monitoring AI agents in customer-facing roles. Officials emphasize that the final approach will reflect Canadian values and feedback.
Experts urge speed rather than perfection. Brian McQuinn, co-director of the Centre for Artificial Intelligence, Data, and Conflict at the University of Regina, described the current moment as an arms race against disinformation actors wielding generative tools. Without swift legislative action, he warned, citizens lose the ability to distinguish synthetic material from authentic human expression.
Teresa Scassa, Canada Research Chair in Information Law and Policy at the University of Ottawa, pointed to the strategic logic of aligning with European standards. She noted that Canada lacks the market heft to compel global tech firms into bespoke compliance regimes. Borrowing from the EU framework offers a pragmatic path, echoing Prime Minister Mark Carney’s vision of middle powers banding together to shape AI governance. The consultation may surface uniquely Canadian nuances around Indigenous data and French language requirements. Still, the underlying message from researchers remains blunt: legislate now, because the technology will not wait for a perfect draft.
