OpenAI’s tool pushes students toward reasoning and alerts parents to self-harm
OpenAI launched Teen Mode in August 2026, a ChatGPT version that automatically enrolls every user under 18. It provides guardrails against destructive content, parent notifications for self-harm concerns, and a prohibition on presenting itself as a friend, claiming personal feelings or suggesting it is anything other than a tool. The launch came a year after a lawsuit alleged that ChatGPT had contributed to the suicide of a 16-year-old named Adam Raine.
Kirman writes that the prohibition on presenting itself as a friend, claiming personal feelings or suggesting it is anything other than a tool is a direct response to the Character.AI wrongful death case, in which a 14-year-old formed an intense attachment to a chatbot. She also writes that a growing recognition that both the content AI delivers and the way AI behaves can cause harm appears to have driven Study Mode’s design.
Study Mode is meant to yield gains that will hold. Kirman cites recent research, without naming it, showing that AI versions that simply do the work for children produce short-term grade improvements that collapse once the tool is removed, leaving students worse off than if they had never used it. OpenAI vice-president for education Leah Belsky calls the feature a transformation from “a straightforward answer engine into a cognitive tutor”. Its Socratic approach refuses answers and requires the student to do much of the thinking herself.
Kirman notes that, for years, the technology industry has contested any causal link between its products and the harms named in a wave of similar lawsuits, each alleging that companies had not merely failed to protect children but deliberately designed products to exploit children. Her reading of Teen Mode runs against that position. Kirman calls Teen Mode “a tacit admission of culpability” and says it reveals what the industry now acknowledges — and does not acknowledge — about its products’ negative effects on young users’ mental health.
As a psychoanalyst who works with young people, Kirman says she has observed up close how social media and AI influence psychological development, affecting not only behavior or mood from day to day but identity formation and character. Her judgment of Teen Mode is qualified: the feature may address some of the harms done to teenagers, she says, while a deeper one remains.
In Kirman’s view, Teen Mode offers something genuinely new because it is accessible to any student with a phone, gently pushes back on easy answers and creates rather than forecloses the discomfort of not-yet-knowing. For a teenager without access to a tutor or a teacher with time for questions, she says a Socratic companion can offer a “real democratization of intellectually rigorous support,” with knowledge and availability beyond what any human source could provide. Where tutors eventually go home and parents become frustrated or occupied, Teen Mode is, in her description, a tireless interlocutor. Structurally, Kirman calls it the “too-good parent,” which is where her objection begins.
Kirman places that concern within broader research on constant connectivity. She cites Jonathan Haidt’s synthesis of smartphone-use data, which she says showed a sharp rise in anxiety, depression and suicidal ideation among teenagers coinciding with technology’s incursion into unstructured, unsupervised time. That incursion, she writes, displaces the private space in which an identity takes shape. She also cites Sherry Turkle’s decades of work on teenagers and technology, including her finding that many young people report not knowing what they think until they post it and see how others respond.
Kirman draws on Donald W. Winnicott, the British pediatrician and psychoanalyst who spent decades studying how a child’s sense of self emerges. He argued that a child needs not perfection but sufficiency from the environment. His best-known “good-enough parent” is good precisely because the parent is not always good: present enough to give the child a secure base, but limited enough — distracted, tired and occasionally wrong — that the child must repeatedly draw on her own resources and find herself capable. Kirman’s too-good-parent analogy remains hypothetical. If such a parent could exist, that parent would spare the child the struggle entirely and the sense of self it builds.
Although Winnicott wrote about early childhood, Kirman says the argument extends into adolescence, when separating from other minds becomes the central task. She credits Peter Blos with mapping that terrain most carefully and says he described adolescence as the work of disengaging from the authority of the minds a teenager has depended on and discovering what is actually hers. A teenager who regards parents as idiots, teachers as frauds and every received opinion as suspect, in Kirman’s account, is doing the important work of clearing ground. She writes: “What makes adolescence generative isn’t that teenagers get it right – it’s that they think they’re right – and that magnificent, insupportable conviction is what gives them the nerve to find out what they actually believe. The adolescent whose intellectual life is always accompanied, always guided, always met with a calibrated response, never gets to become her own authority.”
One counterargument, which Kirman says OpenAI’s designers probably would make, is that Teen Mode is a scaffold rather than a crutch: a teenager who learns to think in dialogue with a rigorous interlocutor will internalize the process and eventually need the tool less. Kirman says that assumption requires an incentive to let go ever arriving. She considers that incentive unlikely because the world is moving rapidly toward greater dependence on AI.
Even if a teenager eventually needed the tool less, Kirman says development is not clearly reversible in the same way: what grows with a scaffold may not be the same as what grows without one. She notes that many technology executives seem to grasp this themselves, sending their children to schools with the resources to enforce a mandate on unassisted work. For a teenager without those advantages, she writes, the default is Teen Mode — or, in her view, something worse: AI without the Socratic guardrails.
Kirman says what she encounters in her office goes further. She encounters young people who lack a reliable internal compass and struggle to know what they feel or want. She argues that this instability is not incidental to the youth mental health crisis but the condition underlying it. A person without an internal compass, she says, is an anxious one: dependent on confirmation, uncertain of her own perceptions and easily destabilized.
Kirman says Teen Mode may ask users to think, but it is still holding their hands, guiding them and keeping dependency going. In her account, that design exacerbates the developmental harm she believes sits at the root of the youth mental health crisis. She calls it the deeper harm the industry and Teen Mode’s designers have not addressed.
Ultimately, she says, a stable sense of self requires moments of genuine trial, including being tested and surviving. The blank page, faced alone, becomes one such crucible, where a person encounters herself in doubt and even dread and also in discovery and joy. Writing in the 1850s, the German philosopher Arthur Schopenhauer observed: “To think for oneself is always a little uncomfortable, and that is why so few people do it.” Kirman warns that people may now be creating a generation for whom that challenge is rarely required. She calls the resulting loss profound, hard to measure and therefore likely to be ignored, and says people come to truly know they have a mind of their own only by passing through the discomfort of thinking by and for themselves. Her concern is not only about cognitive skill, she says, but about losing a reliable sense of their own minds.
Readers in the United States concerned about adolescent mental health or suicide can call or text the 988 Suicide & Crisis Lifeline at 988 or chat at 988lifeline.org. In the United Kingdom, the youth suicide charity Papyrus can be contacted on 0800 068 4141 or by email at [email protected]. Samaritans can be reached on freephone 116 123 in the United Kingdom and Ireland. In Australia, Lifeline can be reached on 13 11 14. Other international helplines can be found at befrienders.org.