Canada_considers_regulating_AI_chatbots_under_online_harms_bill

Canada reconvened its Expert Advisory Group on Online Safety to advise on whether the forthcoming Online Harms Act should explicitly regulate AI chatbots and whether to impose age restrictions on social media and generative agents. The panel of 11 experts is split: three members endorse applying the bill to chatbots, while others warn the law was designed for platforms and not for interactive models. Culture Minister Marc Miller said the government is "very seriously" considering age limits for users under 16.

The core practical challenge is mapping obligations designed for platforms onto AI models and service providers. Key technical and legal vectors practitioners should track include:

• •Definition and scope, whether the law targets host platforms, API providers, model owners, or downstream integrators, and how to treat open-source models.
• •Content liability vs generation, differentiating between moderation obligations for user-posted content and control over model outputs that are generated in real time.
• •Age verification and access controls, the feasibility of reliable age checks without creating privacy or circumvention risks.
• •Transparency and auditing, requirements for model documentation, provenance of training data, model cards, and logs to support enforcement and notice-and-takedown processes.

This is part of a global pattern where national regulators are trying to fold generative AI into existing online-safety regimes. Canada is not alone; the UK and EU have addressed platform harms and the EU AI Act targets high-risk models. Several commentators, including legal scholars like Michael Geist and outlets such as The Hub, argue applying the Online Harms Act wholesale to chatbots risks mismatches: the Act assumes persistent user accounts, content streams, and platform moderation workflows, while chatbot interactions are ephemeral, personalized, and model-driven. Advocacy groups such as Children First Canada and some teens pressed at the party convention support strict age limits, framing youth safety as urgent. Practitioners should expect debates over whether the law will impose technical compliance burdens on model maintainers, require safety-by-design controls, or push companies to adopt restrictive access models and enhanced logging.

If the bill expands to cover chatbots, expect increased compliance overhead for startups and research labs, including mandatory risk assessments, safety testing, and recordkeeping. Potential outcomes include gated APIs requiring identity and age checks, obligations to implement content filters, and formal complaint and redress workflows. Open-source models and pre-release research may face new legal uncertainty if obligations fall on hosts and distributors.

The government will weigh precise definitions, the age threshold, enforcement mechanisms, and whether obligations attach to model creators, API hosts, or platform integrators. Watch for consultation outputs from the advisory group, any draft bill language that clarifies scope, and industry pushback on technical feasibility of age checks and real-time output filtering.

Practitioners need to prepare for compliance engineering work if Ottawa opts to regulate chatbots under the Online Harms Act. The debate exposes a design tension: protect children and curb harms, while avoiding regulatory approaches that are poorly aligned with how generative models are built, deployed, and updated. Expect iterative policy design and rapid technical negotiations between regulators, civil-society advocates, and the AI community.