Canada’s Bill C-34 puts new safety duties on AI chatbot providers

Bill C‑34 is a good example of how “second‑wave” AI regulation is starting to zero in on concrete verticals rather than abstract principles. Instead of trying to govern all AI in one shot, Canada is targeting chatbot providers with a duty of care around harmful content, crisis situations and sycophantic behaviour. That matters because large conversational models like ChatGPT, Claude, Gemini and their successors are the front door through which most citizens experience frontier AI. Rules written here will de facto shape how labs design safety systems across their entire product lines. For the race to AGI, the bill doesn’t change who can train the biggest models, but it does influence deployment strategies and interfaces. If enforced, providers will need more robust red‑teaming, crisis‑response protocols and logging for sensitive interactions. That pushes the ecosystem toward higher fixed compliance costs and could advantage well‑capitalized labs over open‑source communities or smaller startups. It also sets a precedent: once one G7 country frames a “duty to act responsibly” for chatbots, others can copy‑paste and expand it. Over time, this kind of vertical rulemaking may become the dominant way governments try to steer AGI’s social impact without directly slowing research.