"The good of the people is the chief law." — Cicero, De Legibus
There is a tell in technology, and once you learn to read it you will see it everywhere: the louder a company insists on a virtue in its name, the more carefully you should check whether it still practices it.
Consider the word open. In the founding documents of the AI era it meant something specific and generous: open research, open weights, open access, knowledge built in public for the benefit of everyone. It was a promise that the most consequential technology of the century would not be hoarded by whoever got there first. A decade later, the companies that made that promise loudest have become the most closed institutions in the industry. The word did not change. The companies did.
This is not a story about one firm. It is a pattern. And the pattern is worth understanding, because the same move is being run on you right now with other words — safe, democratized, responsible, for everyone.
The Founding Promise
OpenAI was founded in 2015 as a non-profit research lab. Its stated mission was to ensure that artificial general intelligence "benefits all of humanity." Not its shareholders. Not its partners. Humanity. The non-profit structure was the whole point: it was a governance commitment, a way of saying we have built this so that no one person and no one balance sheet can capture it. Early on, the lab published its research and released its models in the open. That openness was the brand. It was in the name.
Hold that founding promise in your mind, because the distance between it and where things landed is the entire story.
How the Door Closed
The retreat happened in stages, each one justified, each one reasonable-sounding, each one moving in the same direction.
First the research got quieter — competitive pressure, safety concerns, the usual reasons. Then the models stopped shipping in the open. Then the corporate structure itself began to migrate away from the non-profit that was supposed to be the entire safeguard. In December 2024 the company announced a restructuring that would convert its core into a conventional for-profit and reduce the non-profit to a minority position. The reaction was immediate. A group of former employees, Nobel laureates, and law professors wrote to the attorneys general of California and Delaware arguing the plan would "subvert OpenAI's charitable purpose" and "eliminate critical governance safeguards."
In May 2025, under that pressure, the company publicly walked it back: the non-profit would retain control, and the business would become a Public Benefit Corporation. A reprieve, it seemed. Then in October 2025 the restructuring went through anyway — the for-profit became a PBC in which the non-profit holds roughly 26 percent, a stake valued around $130 billion against a company valuation that had climbed to $852 billion. Watchdog groups were blunt. Public Citizen called the arrangement, in which the non-profit's mission is effectively subordinated to the for-profit, "impermissible." CalMatters reported the deal with California was, in critics' words, "full of holes." The non-profit is nominally in control. There are, observers noted, numerous ways the for-profit ends up calling the shots.
Read those two numbers next to each other: a mission to benefit all of humanity, and an $852 billion valuation governed by a structure its own founders' generation says no longer protects the mission. That is the arc from open to closed, in ten years, in plain sight.
The Open-Washing Playbook
The word did not retire. It got repurposed. Today "open" is one of the most actively laundered terms in the industry, and the laundering has a name: open-washing — wrapping a closed or restricted product in the language and aesthetics of openness to borrow its credibility.
The clearest case is the "open source" AI model. When a company releases a large model and calls it open source, most people hear free, inspectable, yours to use however you like. What they usually get is an open-weight model: the trained weights are downloadable, but the training data is secret and the license carries restrictions that a genuine open-source license would forbid. The Open Source Initiative — the body that literally maintains the definition of the term — has said plainly that Meta's Llama licenses do not qualify as open source, and accused the company of open-washing. The Llama license has at various points restricted commercial use above a user threshold (Llama 4 permits it only if you have under 700 million monthly active users), discriminated by field of endeavor, and excluded entire jurisdictions. A license that tells you which industries you may operate in and which continent you may live on is many things. "Open" is not one of them.
This matters beyond pedantry. "Open" is a trust signal. People extend goodwill, adoption, and free labor to things they believe are commons. When a company harvests that goodwill while keeping the actual control private, it is collecting the reputational dividend of a commons it has not contributed to. You are meant to feel like a participant. You are a user, on terms you did not write, that can change at the next release.
Why Companies Shed "Open" at Exactly This Size
There is a predictable point in a company's life where openness stops being an asset and becomes a liability to be managed. It arrives the moment the company has more to protect than to gain.
When you are small, openness is your distribution strategy. It earns you contributors, credibility, and a community that builds on top of you for free. When you are enormous, that same openness is just a list of things your competitors can copy and your regulators can scrutinize. The rational move — rational for the balance sheet, anyway — is to keep the vocabulary of openness while quietly retiring the substance. You announce the mission at the founding, when it costs nothing and buys everything, and you renegotiate it at scale, when it costs everything and the goodwill is already banked.
None of this requires a conspiracy. It only requires incentives and time. Which is exactly why you cannot evaluate these companies by their stated values. You have to evaluate them by their structure — by what they have made it possible for themselves to do.
How to Tell Open From "Open"
You do not need to be an engineer or a lawyer to audit a claim of openness. You need three questions, and you should ask them of any tool, model, or platform that markets itself on the word.
Can you leave with your work? Real openness shows up at the exit, not the entrance. If you can export your data, your content, and your audience in a usable form and walk to a competitor, the openness is real. If leaving means abandoning everything you built, the openness was decoration.
Who can change the terms, and how fast? Read who holds the power to alter the license, the price, or the access — and whether they can do it unilaterally at the next version. If the answer is "them, anytime," you are renting, however open the homepage looks.
Does the structure protect the promise, or just describe it? A mission statement is a sentence. A governance structure is a constraint. Ask whether anything actually stops the company from doing the self-interested thing — or whether the only thing standing between you and that outcome is their continued good mood.
The Point
The companies that named themselves after a value did not lie, exactly. They meant it, at the start, when meaning it was cheap. What they did was reserve the right to stop meaning it — and to keep the word anyway, because the word does work that the substance no longer does.
The remedy is not cynicism. It is literacy. When a powerful company tells you what it stands for, treat it as the opening statement, not the verdict. The first to state his case always seems right. Then you check the structure, you check the exit, you check who holds the keys — and you find out what the word is actually doing.
Openness that survives contact with scale is rare and worth everything. Openness that evaporates the moment it becomes inconvenient was never a value. It was a phase of the marketing.
Sources: OpenAI founding mission and corporate history; CNBC and CNN reporting on the May 2025 restructuring reversal; CalMatters (October 2025) on the California restructuring deal; Public Citizen statement on the non-profit's subordination; the ex-employee and Nobel-laureate letter to the California and Delaware attorneys general; Open Source Initiative analysis of the Llama license and "open-washing"; OSI Open Source Definition.
