The federal government is considering extending its planned social media ban for children to AI chatbots, but restricting access to artificial intelligence is proving more complicated than putting minimum age requirements on social media platforms.

Culture Minister Marc Miller has previously said he was “very seriously” mulling a social media ban for kids. Three sources with knowledge of the government’s plans confirmed to The Logic that Ottawa will proceed with that restriction, and is now debating extending it to AI chatbots. At their policy convention last month, members of the Liberal party voted to set the minimum age at 16.

The restrictions would be included in new legislation to regulate online harms that Miller is expected to table this year.

The Logic agreed not to name the sources because they were not authorized to speak on the matter.

Whichever technologies it encompasses, it’s not clear whether the ban would be permanent, or a stopgap until the government institutes a regulator to oversee the platforms. Also unresolved is what types of chatbots would be covered under the proposed restriction.

In a statement, Miller’s spokesperson Hermine Landry said no decisions have been made, but added: “Our government intends to act swiftly to better protect Canadians, especially children.”

If the proposal goes ahead, it would make Canada a pioneer in aggressively trying to restrict children’s access to artificial intelligence tools. Unlike social media bans planned or in place in other countries, restrictions on access to chatbots raises constitutional, privacy and economic questions as Canada makes a push to become a leader in AI.

A blanket chatbot ban for youth would go beyond the age-gate measures Australia enacted earlier this year, for example. That country now requires a wide range of platforms to verify the ages of users trying to access content inappropriate for kids. The rule applies to chatbots, but only those capable of generating sexual or graphic content. 

The U.S. is considering a bipartisan pair of House and Senate bills that would put minimum age limits on “companion” chatbots, which encourage users to have human-like interactions and build relationships with the AI.

The Tumbler Ridge shooting intensified public scrutiny of AI chatbots, after news reports revealed that the B.C. teen who shot and killed eight people in February had troubling conversations with OpenAI’s ChatGPT before the attack. Federal ministers demanded more stringent safety protocols in a series of meetings with the company’s executives, including CEO Sam Altman.

Preventing kids from accessing the tools could prove thorny, though, not least because it runs counter to the government’s plans to foster AI adoption. 

While kids don’t necessarily need Instagram, they do have legitimate uses for AI, including training for the jobs of the future, said Emily Laidlaw, Canada Research Chair in cybersecurity law. 

As a result, a broad prohibition on chatbots for kids is vulnerable to a court challenge, she said. “It’s more difficult to justify banning a general-service AI when there are all kinds of benefits in terms of information, learning and access to knowledge.”  

The government asked Laidlaw to weigh in on the idea as a member of its expert advisory committee on the issue. 

In a speech last weekend in Toronto, Miller said the government would pair any ban with other measures to make platforms safe for kids. “It should not be a partisan thing that you want to protect kids and, once you’ve done that, make sure that platforms have a duty to act responsibly,” he told the crowd at the Global Progress Action Summit. That includes holding platforms and AI companies responsible, he said.

Onstage at that summit, Prime Minister Mark Carney’s wife, the economist Diana Fox Carney, discussed the idea of social media and chatbot bans for kids with Jonathan Haidt, an American social psychologist and author of The Anxious Generation, an influential book about modern technology’s impact on kids. A proponent of social media age limits, Haidt said governments must not respond to AI as slowly as they have to social media, though he stopped short of recommending a chatbot ban.

“If it takes us five or 10 years on AI, it’s going to be a totally different animal,” said Haidt. “We have to speed it up.”

Diana Fox Carney was sharply critical during the session of social media platforms, at one point saying that companies played on the weak parts of human psychology to intentionally get people addicted to their platforms. Video of the event shows the prime minister seated in the front row with his daughter, nodding and taking notes.

Miller’s office would not say when he is expected to table the new legislation.