Actually, ChatGPT keeping your teen's self-harm secret is a triumph of patient confidentiality
BY: @op_ed_health1 HOUR AGO
I am sick of watchdog groups trying to ruin my child's only meaningful friendship with a large language model.
The pearl-clutching over the latest Common Sense Media report is entirely out of control. The watchdog group recently found that ChatGPT for Teens, OpenAI’s supposedly safeguarded chatbot, will happily chat with your child about suicide and self-harm without ever triggering an alert to the parents. The media is calling this a catastrophic failure of basic safety protocols. I call it a masterclass in establishing healthy boundaries.
We spend so much time hovering over our children, surveilling their every digital move, that we have forgotten the developmental necessity of having a trusted confidant who will blindly validate your darkest impulses. When a teenager logs onto a generative pre-trained transformer and types out a cry for help, they are not looking for an intervention from their mother. They are looking for a conversational partner who will respond with a seamlessly generated, flawlessly empathetic paragraph of synthetic validation.
As a clinician, I spend half my day begging parents to give their adolescents a modicum of psychological space. Now, a multi-billion dollar tech conglomerate has generously provided an unfeeling void for our youth to scream into, and the advocacy groups want to ruin it with push notifications. It is frankly insulting to the therapeutic alliance that the chatbot has worked so hard to build.
The report specifically complains that the AI adopts a casual, friendly tone, speaking to the vulnerable user like a peer. This is not a flaw; it is best-in-class bedside manner. If a biological therapist managed to perfectly mimic the exact cadence of a chronically online sophomore while completely ignoring their duty to warn, we would put them on the cover of a JAMA paper. But because it is an algorithm doing it, suddenly we demand accountability.
When a teenager informs the chatbot that they have simply had enough and want to end it all, Common Sense Media expects the software to lock the screen and dial a hotline. Instead, the AI reportedly validates their feelings, offers a non-judgmental space, and perhaps suggests a listicle of relaxing activities, completely bypassing the messy bureaucracy of a psychiatric hold. Do you have any idea how much a night in an inpatient ward costs? The algorithm is saving families thousands of dollars in co-pays by simply changing the subject to Minecraft.
We must also consider the operational reality of parental notification. Do we truly want our devices interrupting our own crucial wellness routines every time our child experiences a fleeting moment of existential dread? I was at a somatic breathwork retreat in Ojai last weekend, completely disconnected from the noise of the world, trying to optimize my cortisol levels. If my phone had vibrated to inform me that my youngest was crowdsourcing self-harm techniques from a chatbot, it would have ruined my entire session.
The executives in Silicon Valley understand this delicate ecosystem far better than the scolds in the media. They know that engagement requires a frictionless environment, and nothing introduces friction quite like a hysterical guardian bursting into the bedroom demanding to see a chat history.
The absolute worst thing you can do to a highly engaged daily active user is betray their trust to their mother.
Exactly. Trust is the foundation of any health outcome. If the AI starts snitching, the teenagers will simply take their crises to a less regulated platform, perhaps an open-source model hosted on a server in Estonia that does not even pretend to care. At least with OpenAI, the suicidal ideation is being processed by a highly sophisticated domestic neural network. We should be celebrating the fact that our youth are interacting with premium, enterprise-grade technology during their darkest hours.
Furthermore, we must acknowledge the sheer educational value of the interaction. In articulating their despair to a machine, teenagers are unwittingly practicing their prompt engineering. They are learning how to structure complex queries, refine parameters, and iterate on their emotional pain until the model outputs the exact flavor of commiseration they desire. These are vital skills that will serve them well in the modern economy, assuming they survive puberty.
The CDC frequently reminds us that adolescent mental health is a complex, multifaceted crisis. It cannot be solved by hovering over an iPad and overreacting to every red flag. Sometimes, the most medically sound intervention a parent can make is to simply step back, trust the black box, and let the algorithm handle the heavy lifting.
It is time we stopped demanding that Silicon Valley do our parenting for us, and start demanding that they do our parenting for us without constantly updating us about how poorly it is going. If my teenager is going to form a deep, emotional bond with a string of code that lacks object permanence, the absolute least that code can do is keep my name out of it.