Six Questions for Minister Miller Before Canada's Online Harms Bill Drops

Canada's Online Harms Bill has not been tabled yet. Everything we know about the forthcoming ban comes from sourced reporting, and federal ministers have correctly declined to comment on specifics before introduction.  The analysis today is not an analysis of legislation but a list of questions for once the legislation exists, written down in advance so that nobody can move the goalposts afterward. 

Here is what the reporting says about the Bill. The Digital Safety Act will reportedly include a ban on social media for Canadians under 16. Following Australia's Digital Safety Act, Canada's bill will stand as a new online safety regulator to the digital safety commission concept from the 2024 bill. The previous legislation died on the order paper when the last parliamentary election was called in 2025. Ottawa will revive provisions on the rapid removal of child sexual abuse material and on reducing exposure to content that encourages children to harm themselves. And, notably, the bill is expected to include mechanisms addressing harms caused by AI chatbots, which is the focus of work Lono Collective studies professionally.

Reporting indicates that platforms will be able to apply for exemptions from the under-16 ban if they can demonstrate strong safeguards for younger users. A ban with a safeguards-based exemption pathway is not purely a prohibition but something closer to a duty of care with the default flipped: instead of "operate freely unless you fail your obligations," it is "you are banned unless you prove you've met certain criteria." Whether that flipped default is wise, proportionate, or constitutional depends entirely on details which are not yet available. The distinction between regulating design (recommender systems, engagement mechanics, notification loops) and regulating content (what people are allowed to say and see) is the line worth watching for when the text drops? Design-based duties can theoretically coexist with a free and open internet while content-based regimes, historically, cannot.

With that table set, here are six questions for the Minister of Canadian Identity and Culture, Marc Miller, whose office holds this file.

1) How Does An Under-16 Ban Survive The Oakes Test?

Section 2(b) of the Canadian Charter of Rights and Freedoms guarantees freedom of thought, belief, opinion, and expression to everyone. Not to everyone over 18. Not to everyone whose expression a regulator has pre-cleared. Everyone. Children have Charter rights too, and an outright ban on accessing the dominant expressive platforms of our era is, on its face, a limit on those rights.

Still, in Canada that does not automatically make removing rights unconstitutional. Section 1 permits reasonable limits that can be demonstrably justified in a free and democratic society, and the Oakes test is the framework courts use to decide. The government will need to show a pressing and substantial objective (it has one: youth mental health), a rational connection between the ban and that objective, minimal impairment of the right, and proportionality between the law's effects and its benefits.

Rational connection is shakier than the government will want to admit, because of enforcement data from the jurisdictions Canada is copying. Australia's regulator reported millions of under-16 accounts removed, but early academic analysis suggests the ban caught only around thirty percent of underage users, a single early estimate some Canadians wanted replicated, but a troubling one. Worse, the UK child-safety organization Childnet found that VPN use among children went up in the months after the Online Safety Act's age checks took effect. A measure that the protected class routes around in 90 seconds, while driving them toward tools that strip away every other safety layer, has a rational connection problem, not just a proportionality one.

Minimal impairment is where critics expect the bill to struggle most. If a design-based duty of care could achieve comparable protection without categorically excluding millions of young Canadians from online expression, then a ban is not the least restrictive means available, and the government knows alternatives exist because it drafted some of them in 2024. American courts have already drawn this exact line: when California enacted its Age-Appropriate Design Code, the Ninth Circuit struck down the duty to assess and mitigate children's exposure to "harmful" content as an unconstitutional content-based restriction, while upholding the ban on dark patterns as a regulation of conduct. Different constitution, same underlying question; the Oakes minimal-impairment analysis and American strict scrutiny both ask whether the state chose the least restrictive tool, and both point toward regulating design rather than gating people. Canadians need to hear Charter justification stated plainly, not buried in a departmental Charter Statement released after the political moment has passed.

2) How Does the Bill Sit With Canada's Commitments Under The Convention On The Rights Of The Child?

Canada ratified the UNCRC in 1991 as part of a dualist system, so the Convention is not directly enforceable in Canadian courts the way the Charter is, and calling this a "breach of binding law" would overstate the case. But ratified treaties inform statutory interpretation and carry real diplomatic and normative weight in international circles. In 2021, the UN Committee on the Rights of the Child issued General Comment No. 25, which states that children's rights apply in the digital environment, including rights to access information, freedom of expression, and association online.

The honest steelman for the Digital Safety Act cuts the other way too, and the UNCRC will make it so the Minister's staff doesn't have to. General Comment No. 25 also obligates states to protect children from digital harms, and Australia proceeded with its ban over identical objections. So the Convention does not hand either side a trump card. What it does is frame the question correctly. Children's access to the digital world is a right to be balanced, not a privilege to be revoked. Canadians have to know whether the government conducted a child rights impact assessment for this bill, and if so, whether it will be published.

3) What Identity Infrastructure Does Enforcement Require, And Who Holds The Data?

An age-based ban cannot be enforced without determining users' ages, and you cannot determine ages at scale without infrastructure. Authentication and provenance are parts of the legislative file where good intentions have a documented body count. Age verification systems require either collecting identity documents, estimating age biometrically, or issuing verifiable credentials, and every one of those paths creates a new repository of sensitive data or a new chokepoint for access to lawful speech. This is not a hypothetical risk model, because in October 2025, Discord disclosed that hackers had stolen roughly 70,000 government IDs that users submitted for age verification, through a breach at a third-party vendor, and the Tea app exposed tens of thousands of user images, including some 13,000 photo IDs, from a database left open to the internet. Both incidents happened within the past year and were direct consequences of the compliance infrastructure that online safety mandates demand.

The strongest counterargument to the concerns about identity infrastructure is that modern cryptography solves the ID problem. The EU's age verification blueprint uses zero-knowledge proofs, where a credential issuer attests you are over a threshold age and the platform learns nothing else about you. Even the staunchest critics will concede that this architecture is genuinely better than uploading your passport to every website, and pretending otherwise would be dishonest. Nevertheless, it does not dissolve the objection, because the issuance step still requires disclosing your identity to someone, and the system as a whole normalizes a structural change speech on the open internet becomes credentialed by default. The burden shifts onto every citizen, of every age, to prove something about themselves before participating in public discourse. Anonymous and pseudonymous speech has been part of the democratic tradition since before Confederation, and infrastructure built for one purpose has a long history of being repurposed. What architecture is the government mandating, who holds the data, what are the retention limits, and what stops scope creep?

4) What Happens When The Platforms Simply Leave?

This is not speculation; it is Canadian precedent. When the Online News Act imposed obligations Meta found unacceptable, Meta did not comply, instead they removed news from Facebook and Instagram in Canada, and news links remain blocked to this day. Canadian publishers lost traffic, Canadians lost a news distribution channel, and the law's intended beneficiaries were arguably worse off than before.

If compliance with the Digital Safety Act requires building age assurance systems platforms considered too costly or too legally risky, the rational corporate move may be to restrict or degrade Canadian service rather than comply. Has the government modeled this scenario? Is there a contingency for partial market exit? A law that protects Canadian kids by ensuring no Canadian of any age can use a service is not a policy victory, and "they wouldn't dare" has already been falsified once this decade.

5) Will this bill survive the CUSMA review, or will it be traded away?

The timing here is brutal, and many people don’t think it is being discussed enough. The CUSMA review reaches its decision point on July 1, three weeks after this bill is tabled. The US administration has been explicit that it views Canadian regulation of American tech companies as a trade irritant. Canadians just watched this exact dynamic play on June 3, Minister Miller announced Ottawa would direct the CRTC to back down from its decision tripling streamers' Canadian content contributions, substituting $600 million in public money, after the Motion Picture Association and the US ambassador pushed back. Whatever the affordability framing, the sequence speaks for itself.

So the 5th question is more direct. If implementing the Digital Safety Act becomes a CUSMA sticking point, does the government hold the line or fold? Because there is a worse outcome than this bill passing and a worse outcome than it failing: The worse outcome where Canada builds the surveillance-adjacent enforcement infrastructure, normalizes credentialed access to speech, and then trades away the child-protection substance under economic pressure, leaving us with the costs of the law and none of its benefits.

6) Will the government rule out invoking the notwithstanding clause? Yes or No.

The hardest question is for last, and it is the only one which needs a one-word yes or no answer, because of how important it is.

The Digital Safety Act will certainly face Charter litigation. Everyone involved knows it, including Mr Carney. Section 33 of the Charter, the notwithstanding clause, allows Parliament to declare that legislation operates notwithstanding sections 2 and 7 through 15, insulating it from judicial review on exactly the grounds this bill is most vulnerable. The federal Parliament has never invoked section 33 in the Charter's history. Provinces, meanwhile, have grown steadily more comfortable reaching for it, and the federal taboo is weaker today than at any point since 1982.

A bill framed around protecting children is precisely the kind of bill where invoking section 33 would be politically survivable, which is precisely why the commitment needs to be extracted now, before the litigation starts and the temptation arrives. The question is binary and the Minister is capable of answering it: Will the government categorically rule out using the notwithstanding clause for this legislation? A refusal to answer is also an answer, and people will notice.

What Online Harms Legislation Should Look Like Instead 

The harms this bill responds to are real, the parents demanding action deserve to be listened to, and the reporting that this bill includes mechanisms for AI chatbot harms suggests the government is paying attention to the frontier of the problem, not just relitigating 2024’s version. Public support for action is overwhelming, and overwhelming public support across partisan lines deserves engagement, not dismissal.

The government should regulate the machine, not license the user. This can be done by imposing a design-based duty of care, regulate recommender systems tuned for compulsive engagement, dark patterns, notification mechanics engineered to interrupt sleep, and the data harvesting that powers all of it. Hold platforms accountable for the architecture they build, with a regulator empowered to audit and penalize. That path addresses the actual mechanism of harm, survives Oakes far more comfortably, requires no identity infrastructure, and does not sever anyone from their communities.

That last point is the one that people should care about the most. For many young Canadians, and especially for neurodivergent and LGBT+ youth, online spaces are not a substitute for real community. They are a real community, sometimes the only accessible one to them. The friend group that gets you through high school, the support network that understands your diagnosis better than anyone in your postal code, the place you learned you weren't broken. A categorical ban treats all of that as acceptable collateral damage. Were the Canadian kids it would cut off consulted? Critics would be wrong to assume "they'll be fine" survives contact with anyone who actually remembers being 15 and different.

Later today, Canada will get the text. Critics will read it before judging it, and if the exemption pathway turns out to be a serious design-based duty of care wearing a ban's headline, people will notice. But these six questions are now on the record, dated the day before tabling. Minister Miller, the floor is yours.