In 2015, a new organization launched with one of the most idealistic missions in tech history: to make sure artificial intelligence benefited all of humanity, not a handful of corporations. It was a nonprofit. Its research would be shared. Its name said the quiet part out loud — OpenAI.
A decade later, that same organization is worth somewhere around $852 billion, has scrapped the cap on how much its investors can earn, keeps its best models firmly closed, and is building toward a future where you pay to access intelligence like a utility.
So what actually happened? A lot of people online have a one-word answer: theft. The real story is messier, more interesting, and — in a couple of ways — more damning than the rant.
Let me walk you through it, because this is one of those cases where the accurate version hits harder than the exaggerated one, precisely because you can’t wave it away.
The Promise
Start with what was actually promised, because you can’t measure a betrayal without it.
OpenAI was founded as a nonprofit by a group that included Sam Altman and — this matters — Elon Musk, who was an early backer and funder. The pitch was explicitly anti-corporate: advanced AI was too important and too dangerous to be owned by any single profit-driven company, so they’d build it in the open, as a public good, to “benefit humanity.” The openness was the whole brand.
For a while, they leaned into it. Early research was published. The culture was academic. If you’d described OpenAI in 2016, “idealistic nonprofit lab” would have been fair.
Hold that image, because almost none of it is still true.
The Drift
The turn didn’t happen in one dramatic moment. It happened in steps, each one defensible on its own — which is exactly how the biggest shifts always happen.
In 2019, OpenAI created a “capped-profit” arm nested under the nonprofit, and took a multi-billion-dollar investment from Microsoft. The logic was practical: training frontier AI costs staggering amounts of money, and you can’t raise that as a pure nonprofit. Fine. But notice what it did — it welded a commercial engine onto a mission that was supposed to be above commerce.
At the same time, the open part quietly died. The flagship models — the GPT series that made OpenAI famous — became closed. You could use them through a paid API, but you couldn’t see inside them. The company that was named for openness had decided its most important work was too valuable, or too dangerous, to be open. Depending on the day, they argued both.
So by the early 2020s, “OpenAI” had become something close to an ironic name. And the biggest moves were still to come.
The Data Question — What’s Alleged, and What Isn’t
Here’s where I have to slow the rant down, because this is the part people get legally and factually wrong, and the wrongness matters.
OpenAI trained its models on enormous amounts of text and images scraped from the open internet — that much is not in dispute, and OpenAI has acknowledged training on publicly available web data. The fight is over what that means.
A growing stack of lawsuits alleges it was theft. The New York Times sued OpenAI and Microsoft in December 2023, claiming millions of its articles were used without permission and that ChatGPT can regurgitate them nearly verbatim, threatening the Times’s business. A consolidated group of authors — including George R.R. Martin, John Grisham, and Sarah Silverman — allege their books were ingested without consent to train the models. As of 2026, these cases are deep in discovery. In one striking development, news publishers have accused OpenAI of stonewalling in that discovery — of telling the court it couldn’t search its own training data when it apparently could.
But — and this is the crucial “but” — no court has ruled that any of this is theft. OpenAI’s defense is fair use: the argument that training a model on copyrighted work to produce something new is legally transformative, the way a person who reads a thousand books to become a writer isn’t “stealing” them. That question is genuinely unsettled law, and it’s being decided right now.
So when someone tells you flatly “he stole all your data,” they’re not stating a fact. They’re stating a lawsuit — one that hasn’t been won. The honest version is heavier, not lighter: OpenAI built a trillion-dollar-scale product on the collective output of millions of people who were never asked and never paid, and whether that was legal is being fought over in federal court as we speak. You don’t need the word “theft” to find that unsettling.
The 2025 Pivot That Dropped the Cap
Now the move that lit the fuse on the current wave of anger, and the one the rant gets closest to right.
In October 2025, OpenAI restructured again. The original nonprofit became the OpenAI Foundation, and the commercial side was reformed into a for-profit Public Benefit Corporation called OpenAI Group. Defenders point out — correctly — that the Foundation still controls the PBC, appoints its board, and holds a financial stake. The nonprofit didn’t vanish. It’s not a clean “nonprofit becomes for-profit” flip.
But here’s the detail that matters most, and it’s the one that should stick with you: they scrapped the profit cap. Under the old 2019 structure, investor returns were capped — famously at 100x. That cap was the last structural remnant of the “we’re not really in this for the money” promise. The 2025 restructure removed it. Microsoft and other investors can now earn unlimited returns. By mid-2026, the whole thing was valued at roughly $852 billion.
A ceiling designed to keep the mission ahead of the money is gone. Whatever you call the structure on paper, that’s the substance.
The public’s writing helped build it. Now the public gets to rent it back.
“Intelligence, Metered”
Which brings us to the stage. Altman has spoken, repeatedly, about a future where AI is a utility — where “intelligence” flows like electricity or water, something you meter and pay for by the unit. It’s often framed optimistically: intelligence so abundant and cheap that everyone has access.
But sit with the shape of that deal, because it’s the emotional core of the backlash and it’s a fair thing to be uneasy about. The training data came, in large part, from the commons — the open web, the free writing, the shared art and code that people put into the world without a paywall. The output is a product you pay a corporation to access. The commons went in; a meter came out.
You can believe, as I do, that OpenAI has built something genuinely astonishing and still find that arrangement worth questioning. Where does the value of collective human output belong — with the company that assembled it into a product, or, in some part, with the millions of people whose work made the product possible? That’s not a rhetorical trap. It’s the actual unanswered question of this decade, and “you’ll pay us by the token” is one very convenient answer to it.
Where I Land
Let me be fair, because fairness is the whole point of a site like this.
Some of the loudest villain narratives don’t survive contact with the facts. It wasn’t one man’s scheme — it was a series of institutional decisions, and one of the founders now cast as a victim, Elon Musk, both funded the thing and then sued it and lost, with a jury ruling in 2026 that he’d waited too long to complain. The nonprofit still legally sits on top. The “theft” is unproven. The closed models and the paid access are, at least partly, defensible responses to genuinely hard problems about safety and cost.
And yet. Strip away the exaggerations and what’s left is still a real story about drift — about an organization that was founded, in writing, to keep transformative AI out of the hands of any single profit-seeking company, and that has become one of the most valuable profit-seeking companies on Earth, its idealistic guardrails quietly removed one by one. The anger online is sloppy with the facts. It is not wrong about the direction.
The Bottom Line
OpenAI didn’t rob a bank. What it did is subtler and, honestly, more important: it took the most idealistic promise in modern tech — this will belong to everyone — and, step by careful step, turned it into a product everyone can pay for. Whether that’s a betrayal or just the unavoidable cost of building something this expensive is a question reasonable people will argue about for years.
But don’t let anyone flatten it in either direction. It’s not simple theft, and it’s not a harmless nonprofit doing its best. It’s the story of what happens when “for humanity” meets a $10-billion training bill — and of who ends up holding the meter when the idealism runs out.
Your work may well be in there somewhere. That’s not a conspiracy theory. It’s just, increasingly, the business model.
