The Globe and Mail reports today that the government will introduce online harms legislation this week that includes a ban on social media for kids under the age of 16. The ban will be framed as a “temporary” measure with the prospect that the can re-establish service after a new digital regulator certifies that they meet its safety standards. I’ve written extensively about why a ban on social media and AI chatbots is a bad policy idea, but it is essential to emphasize that this measure is unlikely to be “temporary.” An age-based ban will require everyone in the country to prove their age before posting a photo on Facebook, uploading a video on TikTok or using an AI chatbot. This raises enormous privacy concerns and turns the government’s AI for All strategy into ID for All.

The government will apparently present this as a temporary safeguard given the prospect of opting back in, but the policy itself requires a regulator and ID for everyone since limiting the services to those over 16 requires knowing the age of all users. This system doesn’t simply disappear since once Canada has required every user to prove who they are in order to police a single age line, that requirement and the infrastructure will remain in place. In other words, the requirement might be reversible, but the data collection and regulatory infrastructure are permanent.

The language of a “kids ban” obscures how the policy actually works. There is no way to keep people under 16 off a platform without determining the age of everyone who uses it, because identifying who falls below the line necessarily means identifying who sits above it. A rule aimed at a minority of users is therefore an age-verification mandate imposed on the entire population, and it makes no difference whether the government calls the result a ban, a restriction, or a safeguard. Tens of millions of Canadians who are not the target of the policy would be required to submit proof of age, typically to foreign third-party verification services, in order to do the ordinary things they already do online. The privacy risks, including the challenge of even applying Canadian privacy law to the collection, are enormous.

None of this would be defensible even if the ban worked, and there is little evidence that it does. Australia’s under 16 ban, the example most cite, has shown how readily children route around age gates through VPNs, borrowed accounts, and false birthdates, with the most at-risk users the most likely to circumvent them. Indeed, even the country’s eSafety Commissioner has acknowledged that the restriction has been difficult to enforce and has yet to demonstrate a measurable reduction in harm. The case for applying any of this to AI chatbots is even weaker.

The forthcoming online harms bill will presumably revive the notion of a duty to act responsibly. That duty, not bans, is what is needed since it could impose requirements on how platforms and AI chatbots are designed, establish transparency obligations, and create liability without locking anyone out or assembling a nationwide ID verification system. The strategy the government launched last week sets out an ambitious vision for AI in Canada. But it cannot deliver that vision by requiring Canadians to prove their age before they can use the tools it is urging them to adopt. That is not AI for all but rather ID for all, and once established, there is no putting the toothpaste back in the tube.