The Skeptic's Case Against Free AI From OpenAI

There's a reflex, when something useful shows up free, to skip straight to gratitude. Free ChatGPT for every verified U.S. K-12 educator through June 2027 — what's not to like? Plenty, says the skeptic, and the skeptic deserves a full and fair hearing rather than a quick dismissal. "Free" is a price, not a promise. It tells you what you pay at the register; it tells you nothing about what you pay everywhere else.

This essay does something the promotional coverage won't: it argues the skeptical side at full strength. Not as a hit piece — the tool is genuinely useful, as earlier essays in this series have said plainly — but because a marketplace of ideas that only prints the optimistic case isn't worth trusting. If the skeptics are wrong, they should be wrong after their best argument has been heard, not before.

Key Takeaways

• "Free" is a business decision, not charity. OpenAI is a company with commercial goals. A free tier for teachers advances those goals; understanding how is just due diligence.
• The data question is real. Even with education-grade privacy defaults, sensitive classroom and student context flows through a third party's systems. That deserves scrutiny, not a shrug.
• Vendor lock-in is a slow cost. Free-until-2027 builds workflows, habits, and dependencies. What happens at the sunset date is a question worth asking now, not later.
• Futurism named the incentive problem. The outlet questioned why OpenAI is pouring money into a major teachers' union — a conflict-of-interest angle the marketing never raises.
• Skepticism isn't rejection. You can use the tool and still insist on hard questions about data, dependency, and motive. Informed adoption beats grateful adoption.

"Free" is a strategy, not a gift

The first move a skeptic makes is the correct one: ask who benefits. OpenAI is not a charity. It's one of the most valuable companies in the world, and it does not give a flagship product to millions of users out of sentiment. There is a business case, and naming it isn't cynicism — it's literacy.

The plausible cases are easy to enumerate. Teachers who adopt the tool for free normalize it for a generation of students, who become paying users or lifelong defaults. A verified educator base is a pristine dataset of high-quality, real-world usage. Goodwill with the teaching profession is politically and commercially valuable when regulation of AI in schools is being written. None of this requires assuming bad faith. It only requires assuming OpenAI behaves like the company it is. This series devotes a whole later essay to the mechanics — follow the money: why OpenAI gives ChatGPT away — but the skeptic's baseline is simply: a free offer this large is a strategy, and you should know the strategy before you commit your classroom to it.

The data question

The second pillar is data. ChatGPT for Teachers ships with education-grade privacy settings — OpenAI states it doesn't train on this data by default — and that's a meaningful improvement over the consumer product. But the skeptic doesn't stop at the default. The question isn't "is the default good?" It's "what is flowing through a third party's infrastructure, and what would it mean if the default changed, the terms updated, or the system were breached?"

Teachers handle sensitive material constantly: student work, behavioral notes, accommodation details, parent communications, occasionally information touching on a child's home situation. Even well-governed, that context moving through an external system is a real exposure surface, and FERPA compliance is a floor, not a ceiling. This series takes the training-data question seriously in its own right — does OpenAI train on your lesson plans? — and the honest answer involves defaults, settings, and trust in a vendor's future conduct. The skeptic's contribution is refusing to treat "not trained on by default" as the end of the conversation. Defaults change. Companies get acquired. Policies get rewritten. The prudent posture is to assume the sensitive stuff needs care regardless of the current setting.

Vendor lock-in and the sunset cliff

The third argument is the slow one, and the easiest to miss because it doesn't bite today. Free-until-June-2027 is long enough to build real dependency. A district that rolls out ChatGPT for Teachers, trains its staff on it, wires it into daily prep, and reorganizes workflows around it has, by 2027, a large sunk investment in one vendor's product — habits, templates, muscle memory, institutional expectation.

Then the free period ends. What happens? Maybe it's extended. Maybe it converts to a paid tier at terms nobody has seen. Maybe the product changes. The point isn't to predict the answer; it's that the skeptic asks the question before the dependency is built, not after. The whole reason the sunset date matters is that switching costs compound quietly. This is the classic shape of platform lock-in: the on-ramp is free and frictionless, and by the time there's a bill, moving off is expensive. A district doing this responsibly plans its exit at the same moment it plans its entry.

The incentive question Futurism raised

The sharpest and most specific skeptical argument doesn't come from us — it comes from Futurism, which questioned why OpenAI is pouring money into one of the country's largest teachers' unions. In a piece headlined You'll Never Guess Why OpenAI Is Pouring Money Into the Second Largest Teachers' Union, the outlet trained its skepticism on the funding relationship behind the union-led AI training effort — the same National Academy for AI Instruction this series covers in its AFT Academy cluster.

The concern, as Futurism framed it, is a conflict of interest: when the vendor whose product teachers might adopt is also helping fund the union body that trains those teachers, the independence of that training is a fair thing to question. A skeptic isn't obligated to conclude the training is compromised — only to notice that the incentive structure isn't neutral, and that "the union endorses it" carries different weight when the vendor helped pay for the union's AI program. That's not a fringe worry. It's exactly the kind of structural question a careful profession should ask out loud, and Futurism deserves credit for putting it on the record while most coverage stayed celebratory.

Skepticism as informed adoption

Here's the turn, and it matters: none of this adds up to "don't use it." It adds up to "use it with your eyes open." The strongest version of the skeptical case doesn't demand rejection — it demands that adoption be informed rather than grateful. Ask what OpenAI gets. Guard the sensitive data regardless of defaults. Plan the exit while building the on-ramp. Weigh union endorsements against who funded the union's program.

There's also a constructive corollary the skeptic tends to land on: dependence on any single vendor's general chat window is the fragile part. Skills and workflows you can carry across models — the kind of portable, purpose-built tooling collected as education agent skills — are a hedge against exactly the lock-in this essay worries about. The healthiest reader of this series holds two thoughts at once: the tool is useful, and none of that usefulness exempts it from hard questions. The next essay turns to the oldest of those questions — academic integrity — which predates ChatGPT but grew with it.

Frequently Asked Questions

Is ChatGPT for Teachers actually free, or is there a catch?

It's genuinely free to use for verified U.S. K-12 educators through June 2027 — no fee at the register. The skeptic's point isn't that there's a hidden charge; it's that "free" advances OpenAI's commercial strategy in other ways (future users, usage data, goodwill), and that understanding those is basic due diligence before committing a classroom to it.

What did Futurism say about OpenAI and the teachers' union?

Futurism published a skeptical piece questioning why OpenAI is funding one of the largest U.S. teachers' unions, headlined around the surprise of that relationship. The concern it raised is a conflict of interest: vendor money flowing to the union body that trains teachers complicates the independence of that training. It's an incentive question, not a proven wrongdoing.

Should I avoid putting student data into it?

Treat sensitive student data with care regardless of the privacy defaults. The product ships with education-grade settings and states it doesn't train on this data by default, which is a real improvement — but defaults can change and systems can be breached. The prudent rule is to minimize what sensitive context you expose, whatever the current setting says.

What's the risk in the 2027 sunset date?

Dependency. Two-plus years of free access is long enough to build workflows, training, and habits around one vendor. When the free period ends, the terms are unknown, and switching costs will have compounded. Planning an exit strategy at the same time you adopt is the skeptic's practical recommendation.

Does being skeptical mean I shouldn't use the tool?

No. The steelmanned skeptical case argues for informed adoption, not rejection. Use the tool where it helps, but ask who benefits, guard your data, plan for lock-in, and weigh endorsements against incentives. Grateful adoption skips those questions; informed adoption answers them.

---

Part 38 of 100 in the ChatGPT for Teachers series. Previously: EdWeek's Verdict: Boon, Bust, or Just Meh?. Next: Cheating Didn't Start With ChatGPT, But It Grew. Browse more builder insights or explore AI skills for education at aiskill.market.