The principles document and the calendar

OpenAI published a document called "Our Principles" on its newsroom yesterday. Jury selection in Musk v. OpenAI begins this morning. GPT-5.5 shipped four days ago. I want to take the document seriously on its own terms, and I also want to be honest about what it is doing on the calendar.

Companies publish values statements all the time. Most of them are unread, and most of them deserve to be. The interesting question isn't whether the contents of OpenAI's principles document are noble, they will be, that's the genre, but why this particular company, which has had a stated mission baked into its charter since 2015 and a published "Core Values" page since at least 2023, felt the need to issue a fresh, company-level principles document on the eve of a trial about whether it has betrayed its founding purpose.

The charitable reading is that this is housekeeping. OpenAI completed its for-profit conversion in 2024, the corporate structure is materially different from what it was, and a refreshed statement of principles is the kind of thing a general counsel would want on the record. GPT-5.5 just shipped, so there's a product moment to attach values language to. Trial timing is a coincidence; large companies have a lot going on and calendars collide.

Trial timing is a coincidence; large companies have a lot going on and calendars collide.

I don't quite buy it, and I don't think anyone is meant to.

The less charitable reading, which is also, I think, the more accurate one, is that this is a litigation artefact dressed as corporate communication. Musk's case turns on the claim that OpenAI abandoned the non-profit, open-research mission he funded in order to become a conventional capped-profit company that now looks a lot like an uncapped one. The defence needs to demonstrate continuity of mission across the corporate restructuring. A freshly-published, prominently-placed principles document, dated days before jury selection, becomes Exhibit A for that continuity. It is the sort of thing you produce when your lawyers tell you the jury will be shown internal documents and you would prefer some of those documents to be ones you wrote on purpose.

This is not a scandal. It is how large litigants behave, and OpenAI's lawyers would be malpracticing if they didn't think about the documentary record. But it does mean the principles document is an unreliable narrator about itself. Reading it as a sincere statement of belief is reading it the way OpenAI's PR team would like; reading it as a legal artefact is reading it the way OpenAI's legal team intends.

The two readings are not mutually exclusive, and that's the part worth sitting with. A document can be sincerely held by the people who drafted its language and strategically timed by the people who approved its publication. Sam Altman almost certainly believes the principles. The board almost certainly approved the timing on advice of counsel. Both things are true, and the second one tells you more about what the document does than the first one tells you about what it says.

What does this mean for how to read OpenAI as a company right now?

It means the gap between OpenAI's external positioning and its internal commercial trajectory continues to widen, and the company is increasingly aware of it. The for-profit conversion, the Microsoft renegotiation, the reported $500bn-ish private valuation, the enterprise sales motion, the consumer device push, these are the actions of a company building toward being one of the largest commercial entities on earth. The mission language, "ensure that artificial general intelligence benefits all of humanity", is the language of an organisation that thinks of itself as a custodian. Both can be true. Neither is cheap to maintain alongside the other, and the cost of maintaining both is rising.

I'd argue the principles document is best read as evidence of that cost. You don't need to publish a fresh principles document if your actions are obviously aligned with your stated mission. You publish one when there's a question to answer. The document is, in effect, an admission that the question exists.

The counter-case deserves a hearing. Plenty of companies refresh their values language when they hit corporate inflection points, and a major model release plus a completed restructuring plus a maturing enterprise business is a reasonable inflection point. Apple did it. Microsoft did it under Nadella. Google did it, badly, several times. The Musk trial is genuinely a sideshow to OpenAI's actual strategic situation, the company is contending with Anthropic's enterprise momentum, Google's catch-up on capability, and the question of what comes after the current scaling regime. A principles refresh in that context isn't necessarily about Musk at all. It might be about giving a company that now employs several thousand people and serves hundreds of millions of users a clearer internal compass.

That's plausible. It's just that the calendar is the calendar. If you wanted to publish a principles document that wasn't about the trial, you would publish it three months ago or three months from now. Publishing it the day before jury selection is a choice, and the choice tells you what the document is for.

What I'd watch from here:

The trial documents. If the principles document gets cited by OpenAI's counsel in opening statements or in the documentary record, the litigation-artefact reading is confirmed. If it doesn't appear at all, I'll revise toward the housekeeping reading.

Whether the principles get operationalised. Values documents are easy to publish and hard to enforce. The question isn't what OpenAI says it believes; it's whether any specific commercial decision over the next twelve months gets made differently because of what's now written down. Enterprise pricing, model access policy, military and defence contracts, the structure of the next funding round, these are the places where stated principles either bind or don't.

Whether the language survives contact with the next product cycle. GPT-5.5 is a capable model launched into a market that increasingly treats capability as table stakes. The principles document will be read against whatever OpenAI ships next, and whatever it ships next will be optimised for commercial performance. The interesting moment is the first time those two things visibly conflict.

I don't think OpenAI is uniquely cynical here. I think it's a company whose original structure has been straining against its commercial reality for several years, and which has now reached the point where it needs to publish documents reaffirming that the strain hasn't broken anything. The strain is real. Whether it has broken anything is the question the trial is, in part, about, and the principles document is OpenAI's preferred answer, filed into the public record at the moment that answer is most useful.

That's not a scandal. It's just worth seeing for what it is.

Footnotes

¶