OpenAI to launch underage ChatGPT with beefier restrictions — but can they guarantee your kid's safety?

The new teen mode is free, and OpenAI says it’s built to be safe even for families who never touch a single parental setting. However, child-safety advocates say some of the most important protections still depend on a parent opting in — and that gap could leave families deciding whether to spend their hard-earned cash closing it themselves.

Here’s what actually changed, where the coverage still has gaps and what filling them might cost.

What does ChatGPT for Teens actually change?

OpenAI told the AP that the teen version, for kids aged 13 to 17, blocks romantic or sexual exchanges and stops the chatbot from suggesting it has feelings — guardrails meant to head off the kind of attachment a teen might form with the tool. It also restricts how the chatbot responds to topics like suicide, self-harm and eating disorders, while shifting homework help towards guiding a student to formulate an answer themselves rather than simply producing one.

OpenAI doesn’t perform age verification. Instead, it estimates whether the individual is a minor from how they use the app, automatically shifting anyone it flags into the teen version. Parents and teens can also link accounts, which unlocks quiet hours and sends notifications tied to eating disorder-related conversations or high-risk situations, such as signs a teen may be exhibiting signs of hurting themself.

Why ‘free’ protection may not be complete

Linking accounts requires both the teen and the parent to opt in, and some of the most consequential safeguards, including limits on how much the chatbot remembers about a teen over time, are only available through that setup rather than being switched on by default.

Brendan Bouffard, associate general counsel for AI and privacy at children’s advocacy group Fairplay, cautioned against “parents being lured into this false sense of safety,” in comments to the AP, arguing the default protections leave a real hole for families who assume the basic version is enough on its own.

That’s the financial decision that looms with Canadians: Whether to treat OpenAI’s free safeguards as sufficient, or budget for a second layer of oversight.

What would it cost to close the gap?

Third-party monitoring apps are the main paid alternative, and pricing depends on coverage. Qustodio’s basic plan runs about US$43 a year for time limits, web filtering and location tracking, while its more complete tier, with AI-powered alerts and social monitoring, runs between US$89 and US$138 a year depending on how many devices are covered.

For a Canadian family with more than one teen or device, that can mean roughly C$60 to C$190 or more a year just to backstop features OpenAI already offers for free through parental controls, if both the parent and teen agree to turn them on.

What Canadian parents should do first

Before spending anything, link accounts inside ChatGPT itself; it costs nothing and closes the biggest gap advocates flagged. Turn on quiet hours if screen time is a concern, and check the notification settings so alerts about self-harm or eating-disorder content actually reach a parent’s phone. Then have a direct conversation with a teen about how the chatbot gets used for socialization, not just homework, since that’s the use case the free settings can’t fully police.

Only after that conversation does a paid monitoring app start to make sense, and only for the specific gap it fills, such as broader web and social monitoring the built-in tools don’t cover.