The Canadian government has a two-pronged legal strategy to counter rising hate: New Criminal Code provisions to address offline harms, such as intimidation or protests outside schools and places of worship, and the digital safety bill introduced in June, which targets online harms. A new legal duty to act responsibly, which would require internet platforms to publish digital safety plans, provide greater transparency on their efforts to counter harms, and occasionally remove content, serves as the foundation for fighting back against the spread online of illegal content such as non-consensual distribution of intimate images, cyberbullying, terrorist content, and content that incites violence or hatred.
These are worthwhile measures. But for those concerned with antisemitism, Islamophobia, or other forms of hate, my Globe and Mail op-ed notes that stricter rules against illegal content are basic table stakes. The bigger challenge is the far larger category of content best described as “awful but lawful”: the tropes, conspiracy theories, and coded dehumanization that cause serious harm but remain constitutionally protected expression. A law mandating this material’s removal would invite an immediate Charter challenge. Instead, we need to figure out how to address harms without overstepping constitutional safeguards for freedom of expression.
Doing nothing is not an option; the harms are real. Antisemitism offers the starkest current example. CyberWell, a non-profit that tracks antisemitic content online, documented an 86-per-cent spike in the three weeks after the terror attacks on Oct. 7, 2023, and researchers have found statistically significant correlations between concentrations of antisemitic posts and the locations of real-world attacks on synagogues.
The conventional answer has long been that the best remedy for bad speech is good speech. Yet that approach may no longer be effective, as engagement-driven algorithms give conspiratorial content a built-in distribution advantage and online echo chambers or “filter bubbles” mean that counter-speech rarely breaks through. There is, however, a potential solution hiding in plain sight: The platforms themselves say that much of the awful but lawful content violates their terms and conditions.
For example, Meta prohibits dehumanizing speech and, after consulting 145 external stakeholders, extended its rules to attacks using “Zionist” as a proxy for Jews. TikTok bans Holocaust denial and YouTube prohibits content promoting hatred based on religion or ethnicity. These policies are not a function of government legislation, but rather result from voluntary commitments adopted by the companies themselves.
The policies may say many of the right things, but the enforcement record to date tells a different story. CyberWell’s 2025 report found that nearly half of the antisemitic content that violated platform policies remained online even after the companies were notified by those tasked with identifying offending content. When Meta scaled back proactive enforcement in January, 2025, ADL researchers documented a nearly fivefold increase in antisemitic comments on the Facebook pages of Jewish members of the U.S. Congress, even though the company’s written policies remained unchanged. The content can even be profitable, with platforms and social media influencers alike earning revenue from posts that the platforms’ own rules prohibit.
One way to bridge the disconnect between policy promises and enforcement failures would be to hold platforms legally responsible for their own policies. For decades, the U.S. Federal Trade Commission has treated privacy policies as enforceable commitments, creating privacy obligations premised not on government-mandated rules but on companies’ own undertakings. The effect is to create binding privacy obligations backed by real enforcement. Canada has something similar, with the Competition Bureau using the same reasoning in 2020 to reach a $9-million settlement with Facebook over misleading privacy claims.
Whether through the new Digital Safety Commission or the Bureau’s existing deceptive marketing framework, a regulator would audit platform conduct and impose potential liability for systemic failures to comply with the company’s existing content policies. There would not be liability for individual moderation errors, and the government would stay out of regulating awful but lawful speech. Instead, the companies themselves would be held to account for their own promises to address the harms on their platforms. They could respond by promising less, but any retreat would at least be visible, giving users and advertisers a clear sense of where they stand.
The reality is that no legislation can fully address online harms. But holding platforms to their own commitments offers what years of debates have failed to produce: A way to address harmful but lawful content while limiting government intrusion into speech regulation. The platforms have told Canadians, in their own published rules, that this content has no place on their services. The government should use the digital safety bill to make sure they keep their word.
