Josh Dehaas: Carney’s plot to censor the entire internet

Fresh fruits and vegetables formed the backdrop and the framing for Prime Minister Mark Carney’s announcement about a Canadian food security program at the Ontario Food Terminal in Toronto, Ontario on Thursday, June 11, 2026. (Peter Power/Postmedia News)

Article content

The Carney government signalled for months that it would pass a law to protect kids from online harms, including by banning access to social media for those under age 16. What we got instead may be the biggest threat to free expression in decades: mandatory censorship of social media and AI.

Article content

The Safe Social Media Act, Bill C-34, aims to keep kids from signing up for TikTok and Instagram, which is bad enough since it means teenagers will be censored. While many parents don’t want their teens on TikTok, others believe their 15-year-olds are mature enough.

Article content

Article content

Article content

Still, the much bigger concern is that the act would create a new Digital Safety Commission with the power to force social media companies and artificial intelligence to censor the speech of adult Canadians.

Article content

Article content

The Digital Safety Commission would enforce “duties to act responsibly” by “mitigating the risk” that Canadians are exposed to “harmful content.” Mitigating risk means exactly how it sounds: censoring as much content, either by blocking it or demoting it, as the regulator requires.

Article content

Social media content is by definition expression and protected by the Canadian Charter of Rights and Freedoms. The outputs of artificial intelligence chatbots are also protected speech. They are, after all, nothing more than a reflection of human-created speech and writing.

Article content

The government can only limit expression if the limit is demonstrably justified in a free and democratic society. Some of the limits proposed in C-34 easily pass this test, including requiring social media companies to take down intimate images communicated without consent or child abuse images, and blocking AI from telling its users to commit crimes or to engage in self-harm.

Article content

Article content

But the “duty to act responsibly” also includes much more subjective speech like content that “foments hatred.” Plenty of content that one person (or AI chatbot) may believe is “hatred” is another person’s reasoned argument. It must not be censored.

Article content

Social media companies and AI companies will be forced to err on the side of caution, taking down anything that comes close, or risk fines up to $10 million or three per cent of global revenues. This will mean a sanitized version of the Internet where we won’t be able to debate difficult topics or get to the actual truth.

Article content

Just consider what the Supreme Court has said counts as “hatred” in the context of criminal hate speech and anti-discrimination law. The court has said hatred is only extreme vilification and detestation. But what exactly is vilification or detestation? We’re told to look at the “hallmarks of hatred.” These are extremely fuzzy and cover many things free people need to be allowed to say.

Article content

The “hallmarks” include speech that calls a group “animals” or “alleges a targeted group is plotting to destroy Western civilization.” This can be hateful. But what if some group of people is plotting to destroy Western civilization? X would be required to filter such accusations out of the conversation. ChatGPT would be required to hide or contradict these claims.