You Told It Everything

What a free chatbot therapist and a company that paid teenagers twenty dollars for their whole phone reveal about the trade you make every time something is free.

A fifth comforting myth about privacy hides in plain sight because it never feels like a myth. It feels like a bargain. The first four myths were about other people exposing you: what institutions recorded, what your tools failed to hide, who was watching, who you were connected to. This one is about a choice you make yourself, willingly, dozens of times a week, usually while feeling clever for getting something for nothing. The myth is that "free" is free—that the app costing you no money costs you nothing, that the frictionless feature is a gift rather than a transaction. It is always a transaction. When the product is free, you are the price, and data is the currency. The only real question is whether you can see the bill.

The clearest proof isn't a hack or a breach. It is two ordinary scenes: a person late at night typing their most private fears into a free chatbot, and a teenager installing an app for twenty dollars a month. Both are voluntary. Both feel harmless. Both are the whole lesson.

THE CORE IDEA. Convenience and security are two ends of one dial, and "free" is just the setting where you pay in data instead of dollars. Every free feature is a purchase, and the price is privacy — usually the most intimate privacy you have. Privacy is not refusing the trade. It is knowing the price before you pay it.

You told it everything.

Start with the most intimate version, because it is happening right now, at a scale no surveillance program ever achieved: people are confessing the deepest facts of their lives to free AI chatbots, and they are doing it precisely because it is easy, always available, and feels private.

In July 2025, on a podcast, the person best positioned to know said the quiet part out loud. Sam Altman, OpenAI's CEO, warned that conversations with ChatGPT carry no legal confidentiality whatsoever. "People talk about the most personal [stuff] in their lives to ChatGPT," he said. "People use it — young people, especially, use it — as a therapist, a life coach." And then the catch: "If you talk to a therapist or a lawyer or a doctor about those problems, there's legal privilege for it... And we haven't figured that out yet for when you talk to ChatGPT." In plain terms, he added, if you are involved in a lawsuit, "OpenAI would be legally required to produce those conversations today."

Sit with that. The single most private confessional most people have ever used — more candid than what they tell a spouse, because it feels like typing into a void — has none of the protections of the professionals it imitates. A conversation with a therapist is shielded by law. The same words typed into a free chatbot are a discoverable business record. And this is not hypothetical: OpenAI has been under a court order in its litigation with the New York Times to preserve the chats of hundreds of millions of users, including conversations users believed they had deleted. The delete button was a convenience. The retention was the reality.

Nobody was forced into this. Hundreds of millions of people chose the free chatbot over the four-hundred-dollar therapy session, the free legal question over the lawyer's retainer, the free 3 a.m. confidant over the friend they didn't want to wake — and every one of those was a rational, understandable trade of privacy for convenience and cost. The trade is not stupid. It is just invisible, because the price is deferred, abstract, and collected somewhere you will never watch it happen.

KEY TAKEAWAY. The free tool that feels the most private is often the least legally protected. Feeling private and being private are unrelated properties, and the gap between them is exactly where the price is hidden. What you tell a professional is privileged; what you tell a free app is evidence.

They put a number on it.

If the chatbot shows the trade at its most intimate, a different case shows it at its most naked, because someone was willing to state the price in dollars — and people took the deal, including children.

From 2016 until it was exposed in January 2019, Facebook ran an internal program called Project Atlas. It paid people between the ages of 13 and 35 about twenty dollars a month, plus referral fees, to install a "Research" app on their phones. To hide its own involvement, Facebook distributed it through third-party beta-testing services. What the app installed was a root-level certificate and a VPN — the deepest possible access to a device — and what that access delivered back to Facebook was, in the reporting's own list, "private messages in social media apps, chats from instant messaging apps, including photos and videos sent to others, emails, web searches, web browsing activity, and even ongoing location information." It could see data "even where the app uses encryption." Not a slice of a phone. The whole phone.

The detail that should stop you is the paperwork for the minors. Teenagers aged 13 to 17 could enroll with parental consent, and the consent form parents signed stated, in flat administrative prose, that "the inherent nature of the project involves the tracking of personal information via your child's use of apps." A parent read that sentence, decided twenty dollars was worth it, and signed. The teenager decided a twenty-dollar gift card was worth total surveillance, and installed it. This is the convenience-security trade with the mask off: a company named a price for a human being's entire digital life, and the answer, often enough to make it a program, was yes. Apple was angry enough to revoke Facebook's developer certificates over it — not because the trade was illegal, but because Facebook had smuggled it past the rules. The trade itself was perfectly legal. It usually is.

Why are the young fluent in a language they can't read.

The thread connecting these two cases is not an accident of my choosing. It is the same population in both: Altman said young people especially use ChatGPT as a therapist, and Facebook's price list started at thirteen. That is worth understanding, because it is the clearest window into where all of this is going.

Young people are the most fluent data-for-features traders alive. They will hand over a contact list for a filter, a location for a streak, a face for a fun aging effect, a diary's worth of feeling for a chatbot that answers at 3 a.m. This is not because they are careless; it is because they are native to an economy where data is the ordinary medium of exchange, as invisible and unremarkable to them as air. And that fluency is exactly the problem, because being fluent in spending a currency is not the same as being able to read the price. A generation that trades data reflexively, for features that are genuinely delightful, is a generation making enormous, permanent purchases without ever seeing a total — and the records they are generating now, the confessions and the locations and the messages, will still exist, still be discoverable, still be for sale, decades after the filter that bought them is forgotten.

THE ONE-LINE LESSON. When something is free, you are not the customer, and you are not even the product — you are the supplier, paying in data you will never get back, for a convenience you will forget by next week.

The mirror: the call is coming from inside the house.

This is where the case completes the series, turning the threat around. In the four cases before it, the danger came from somewhere else. The Skripal officers were exposed by institutions that had recorded them. He and Wang were beaten by a compromised counterparty. Surveillance Watch mapped an industry aimed at you. The Golden State Killer was undone by his own network. In every one, an outside force did the exposing, and you could point at it.

Here there is no one to point at but yourself. No one hacked the person confessing to the chatbot; they typed it in. No one stole the teenager's phone data; he installed the app and cashed the gift card. The exposure in this case is authored, keystroke by keystroke and tap by tap, by your own entirely reasonable preference for things that are free, easy, and immediately gratifying. That is what makes it the hardest of the five to defend against and the most relentless. You can learn to distrust an institution, a tool, an industry, a network. It is much harder to distrust your own desire for the free version, because it never presents itself as a risk. It presents itself as a deal.

What free is actually worth.

Honesty requires the counterweight, because the wrong lesson here is the one that makes people quit. Free tools aren't a scam, and the answer isn't to pay for everything or use nothing. Much of what you get for your data is genuinely valuable, and a great deal of the data you trade genuinely does not matter. A free maps app that knows your commute is a fair trade for most people. A free game that sees you play it is nobody's crisis. The problem is not the existence of the trade; it is making every trade unconsciously, at the same flat exchange rate, so that the confession you would never put in writing and the trivia you would shout in the street are handed over with identical ease.

The skill, then, is not refusal. It is triage — knowing which data is cheap and which is irreplaceable, and spending accordingly.

THE MINDSET SHIFT. Stop asking "is it free?" and start asking "if it's free, what am I paying, and is this a thing I can afford to spend?" Some of your data is small change. Some of it — your health, your fears, your children's whole digital lives, the thoughts you'd never say aloud — is the family silver. Stop paying for conveniences with the silver.

What to actually do with this

The habits that follow are less a checklist than a pricing instinct.

Assume free means paid in data, and read every "free" as "in exchange for." This single reframe does most of the work. The maps app, the chatbot, the filter, the loyalty program, the research gig that pays twenty dollars — each is a price tag written in a currency you're trained not to see. Just noticing the transaction is most of the defense.

Keep the family silver out of systems you don't control. Your most sensitive thoughts — medical, legal, financial, the genuinely private — do not belong in a free consumer chatbot that retains them and can be compelled to produce them. For those, use tools with real confidentiality: an actual professional where privilege applies, an enterprise or paid tier with contractual data protections, or a model that runs locally and keeps nothing. The convenience of the free confessional is real; it is just not worth your most discoverable secrets.

Teach the young people in your life to read the price, not just spend it. They are the most fluent traders and the least equipped to see the total. The lesson is not "don't use these things" — that fails instantly — but "everything free is charging you something; here is how to see what." A thirteen-year-old who understands that a gift card bought his whole phone has learned the most valuable privacy lesson.

And change the defaults, because the default is the trade someone set for their benefit, not yours. When anything is free, its settings are tuned to extract the most data the maker could get away with. Spend sixty seconds turning off sharing, the training-on-your-data option, and background collection. The free tool will usually still work; it just works a little more in your favor.

The comforting version of privacy says: take the free thing; it costs you nothing. These cases say something harder and truer. Free was never free. There is only the price you paid knowing, and the price that was quietly deducted from a part of your life you'll wish you had guarded — the confession you typed into the void, the phone a teenager sold for twenty dollars, the record that outlives every reason you had for creating it. You told it everything. The least you can do now is decide, next time, what everything is worth.

Sources and further reading

The confessional — AI chatbots and the privilege gap

• Sam Altman warns there's no legal confidentiality when using ChatGPT as a therapist — TechCrunch, on the July 2025 remarks and the litigation-driven retention of user chats.
• ChatGPT Has No Legal Privilege – Is This a Problem? — Artificial Lawyer, on what the absence of privilege means in practice.
• Altman: anything you say to ChatGPT can be used against you — Yahoo Tech.

The price tag — Facebook's Project Atlas

• Facebook pays teens to install a VPN that spies on them — TechCrunch's original investigation, with the $20/month figure, the 13–35 age range, and the root-certificate access.
• Facebook Paid Teens $20 to Install a 'Research' App That Collects Private Data — The Hacker News.
• Apple blocks Facebook's Research VPN — TechCrunch, on Apple revoking the enterprise certificate.

Fifth in the Sparkforge digital-privacy series, following the Skripal, He & Wang, Surveillance Watch, and Golden State Killer pieces. Case facts are drawn from the linked reporting; quotations reflect those sources as published.