Commenting on the launch of ChatGPT for teens

I was interviewed by Élyse Allard for a segment on Ça Nous Regarde, hosted by Madeleine Blais-Morin on Radio-Canada’s Première Chaîne, discussing OpenAI’s rollout of dedicated features and guardrails for teenagers. The initiative was framed around curbing academic cheating, introducing parental controls, limiting usage hours, and suppressing expressions of romantic attachment or sensitive mental health topics amid mounting legal scrutiny and lawsuits.

In the broadcast clip, I highlighted a fundamental engineering reality: truly bulletproof guardrails remain an illusion. Because conversational models possess near-human linguistic versatility, motivated users can easily co-create coded languages or adversarial prompts with the system to route around rule-based restrictions.

In our broader conversation for the piece, I also reflected on the systemic dynamics at play:

Links