Artificial intelligence
What I Want My Kid to Know About AI Before They Leave for College
A letter to parents of rising freshmen
If you have a kid leaving for college in the next year or two, you already have a list. Sheets, mattress topper, fan, surge protector, the bank account, the renter’s insurance, the conversation about laundry, the conversation about consent, the conversation about not posting your address online. The list is long and the time is short and most of it gets done by August.
There’s one conversation that doesn’t tend to be on the list, and I think it belongs near the top: the one about artificial intelligence. I’m going to walk you through it the way I’d want a friend to walk me through it.
(There’s also a small stack of legal and medical paperwork worth sorting out around an eighteenth birthday — the day your child becomes an adult, the rules about who can be told what change overnight. That’s a conversation for a good attorney, not for me. I’m here for the other one.)
What your kid is actually walking into
Most parents have not yet had this conversation with themselves, so let me describe what your child is walking into, plainly, and then suggest what to say to them.
Your kid is about to live in close daily contact with a kind of software that didn’t exist when you went to school. It will be their roommate, in the sense that they will talk to it more hours per day than they talk to almost any human. It will be their tutor — or their ghostwriter, depending on the choice they make. It will be their first stop for embarrassing questions, romantic questions, mental health questions, questions about their own identity. It will, in some cases, be their therapist.
It will also be, in ways they may not notice, retentive. Every message they send may become training data. Every photo they upload may be analyzed. Every voice memo may be indexed. The legal landscape is changing fast, and the defaults are not in their favor.
This is not a panic. It is a new domain that the people we love are entering, and like every new domain, there are three things worth talking through before they go.
Privacy: what never to put in a chatbot
There is a small list of things your child should commit to keeping out of any AI tool, no matter how casual the relationship feels. Social security number. Bank account numbers. Photographs of legal documents. Mental health crises, attached to their real name. Drug use, attached to their real name. Detailed relationship information that names other real people. Anything they wouldn’t say in front of a stranger sitting on a bus next to them — because in a meaningful sense, that is the audience.
The argument is not that AI is malicious. The argument is that AI is durable. Things said to a person are forgotten. Things said to a system tend not to be.
Oracle versus tutor
The single most important distinction your child can learn, and the one most college students get wrong for the first semester, is the difference between using AI as an oracle and using AI as a tutor.
Oracle use looks like this: “Write my essay on the French Revolution.”
Tutor use looks like this: “Help me understand the difference between the moderate and radical phases of the French Revolution, then quiz me on the distinction, then watch me draft an essay and tell me where my thinking is weakest.”
The first costs the same as the second. The second produces a student who can think. The first produces a graduate who cannot. Your child will be tempted by the first because everyone is tempted by the first. The conversation worth having is about why the second is the trade that pays off. This is the heart of what I teach, and the people I teach are not only the eighteen-year-olds — it’s the parents, too.
The face and voice problem
Your child’s phone contains hours of audio and video of them. Most of it lives on social media, where the privacy settings only matter until someone they know reposts it. The phrase “deepfake” is a few years old; the technology is now ordinary. A thirty-second clip of audio is enough to clone a voice. A handful of photographs is enough to produce a realistic video.
There are two consequences worth naming.
The first is that AI-generated impersonation is already being used to scam family members. (I’ve written about that separately, in When AI Sounds Like Family.) Your child should know about a Verified Family Code — a spoken phrase you agree on together, never written down, used to verify any urgent phone call.
The second is harder. AI-generated intimate imagery — sometimes generated from ordinary public photos, sometimes generated in collaboration with a malicious peer — has become one of the fastest-growing forms of harm against teens and young adults. Your child should know that if it happens to them or someone they know, the right move is to come home — to a parent, a counselor, the police, an attorney — and not try to handle it alone. Quiet, immediate adult help is the difference between an incident and a catastrophe.
The conversation
You can have this conversation in twenty minutes, on a drive, with no slides. You don’t need to be an expert in any of it. You just need to have thought about it once before they go — which, if you’ve read this far, you now have.
I taught the first generation of law students how to do their research online, back when the internet was the strange new tool nobody’s parents understood. The job hasn’t really changed. A new tool arrives that’s powerful and a little dangerous, and the people who do well with it are the ones who were handed a few honest words about it early, by someone who wasn’t trying to scare them. That’s all this is. Be that someone for your kid.