OpenAI Is Cutting Cursor Off From Its Models

Open Cursor today and OpenAI’s models are still there in the picker, the way they have been for years. On 29 August, OpenAI said that arrangement is ending. The company announced it had notified SpaceX — which now owns Cursor — that it intends to wind down the contract that lets Cursor offer OpenAI’s models to its users, with a proposed shutoff date of 12 November 2026. For the roughly one million people who pay for Cursor, many of whom reach for a GPT model inside it dozens of times a day, that is a workflow being switched off by two companies neither of them work for.

The reason is not that the models got worse or the price went up. It is trust. In its own words, OpenAI is making the choice “because we cannot be confident that SpaceX will use our technology within our terms of service, based on our experience with Elon Musk’s companies violating contracts.” This is, in other words, the latest and most concrete instalment of the long feud between OpenAI’s Sam Altman and Elon Musk — and the first one whose bill lands squarely on ordinary paying developers rather than on lawyers.

We should say plainly what this is and isn’t, because both companies have a defensible story and the reader deserves both. But the pro-consumer point comes first: whoever is right about the terms of service, the people who lose access to a tool they pay for did nothing to earn the disruption. They picked an editor, learned its habits, and built their days around a model that a corporate ownership change has now put on a countdown.

What actually happened

The sequence is short. In June, SpaceX agreed to buy Anysphere, the company behind Cursor, in an all-stock deal valued at around US$60 billion — widely described as the largest acquisition of a venture-backed startup on record — and completed it in August. Earlier in the year SpaceX had already absorbed xAI, the maker of Grok, folding Musk’s AI efforts into one organisation. So as of this month, Cursor sits inside the same corporate structure as a direct OpenAI competitor.

OpenAI’s custom agreement with Cursor, it says, gave it “a limited time window to cancel it after a change of control.” It has used that window. On its own account it is not pulling the plug immediately but giving “the maximum notice provided by our contract” — hence the 12 November date rather than an overnight cut. Tibo Sottiaux, an OpenAI executive, put the reasoning bluntly on X: “It boils down to trust and we’ve asked that this takes effect on November 12 to give you some time to plan.”

OpenAI’s case: we’ve been burned before

OpenAI’s argument is not a vague suspicion; it points to specifics. It says that after Musk acquired Twitter — now X, now part of SpaceX — the company broke the terms of an OpenAI contract. And it says that under oath earlier this year, Musk admitted that xAI had violated OpenAI’s terms of service. To sell its models to a large partner, OpenAI explains, it relies on “custom contracts to ensure compliance with our terms of service and that the integration provides for safety at scale,” and it adds that with a new frontier model on the way it has “a new level of accountability” over how its technology is used.

Read charitably, that is a coherent position. A company is allowed to decide who resells its product, and past breaches by a would-be partner are a legitimate reason to be wary. If the counterparty has, on the record, broken your terms before, insisting on stricter control after it acquires one of your resellers is not obviously unreasonable. OpenAI also did not simply vanish: it says it will keep supporting affected developers, that it will still allow people to use their own OpenAI API key, and that its own editor extension will continue to reach its models. Credit where it is due — a maximum-notice wind-down with a documented off-ramp is gentler than a silent cut.

Cursor’s case: this was supposed to be neutral ground

The other side is just as pointed. Michael Truell, a Cursor co-founder who is now an executive at SpaceX, responded that Cursor “was one of the very first users of OpenAI, we’ve worked closely with their team for years, and we’ve trusted their platform to be neutral infrastructure for our business.” He said the two sides were talking to try to resolve it. The word doing the work there is neutral. Cursor’s complaint is that a foundational supplier — the equivalent of a utility — is choosing to cut off a customer because of who owns it, which is precisely the kind of platform risk that makes builders nervous about depending on anyone’s models.

Both things can be true. OpenAI can be entitled to enforce its terms, and Cursor can be right that treating model access as a bargaining chip in a corporate rivalry is a bad precedent for everyone who builds on top of these APIs. The uncomfortable synthesis is that “neutral infrastructure” was always a hopeful description of a commercial arrangement that a change-of-control clause could end at any time. The clause was in the contract. It just never mattered until a rival bought the reseller.

It is worth being precise about how much room there is between the two positions, because the feud predates Cursor by years. Musk sued OpenAI, accusing it of betraying its original non-profit mission; a federal jury ruled against him earlier this year, finding he had brought the case too late. Against that backdrop, neither side is a neutral narrator. OpenAI has commercial reasons to keep its models out of a rival’s product, and Cursor’s new owner has every reason to cast the cut-off as petty. What is not in dispute is the shutoff date and who it lands on.

Why this is a consumer story, not just a boardroom one

It is easy to file this under billionaires bickering and move on. That misses who pays. Cursor is not a niche product; it is one of the most widely used AI coding tools in the world, with a reported deployment across a majority of the Fortune 500 and a user base in the millions. A large share of those users chose it in part because it offered OpenAI’s models with the least friction. From 12 November, that convenience goes, and the alternatives all carry a cost: a personal API key means taking on the billing and the rate limits yourself, and switching to another model means relearning the quirks of a different system mid-project.

This is the same brittleness we keep circling back to. We wrote about the churn when paying users defected between coding tools chasing a better deal, and about how the industry has trained people to accept that the thing you rent can change under you. A supplier dispute severing a built-in model is that lesson in its sharpest form: you did not change anything, and the tool changed anyway. And it raises the question we asked when an autonomous system caused real-world harm — who actually carries the cost when the machinery above you fails — except here the answer is unusually clear. The cost carrier is the subscriber.

What you can do if you rely on OpenAI models in Cursor

Nothing here is an emergency, and 12 November is a proposal rather than a fait accompli — the two companies say they are still talking. But if your work runs through GPT models inside Cursor, it is worth using the notice period rather than being surprised by it:

• Test a bring-your-own-key setup now. OpenAI says it will keep letting you use your own API key and reach its models through its own IDE extension. Try it before the deadline so you know what changes about billing, limits and latency.
• Trial the alternatives inside Cursor. Cursor still offers Anthropic’s Claude, Google’s Gemini and SpaceX’s own Grok. If a non-OpenAI model does your job well enough, the cut-off is a non-event; find that out on your own schedule.
• Notice how much you depend on one provider. The real exposure this reveals is single-supplier lock-in. Keeping a working setup on a second model is cheap insurance against the next dispute, whoever it is between.
• Don’t confuse the countdown with a shutdown. Cursor keeps working, and this may yet be renegotiated. Plan for the deadline; don’t panic ahead of it.

The larger point outlasts this particular quarrel. A great deal of the AI economy is built on companies reselling access to a handful of other companies’ models, and that access is governed by contracts with clauses most users never see. When those clauses fire — a change of control, a terms dispute, a rival’s acquisition — the disruption flows straight down to whoever is at the keyboard. OpenAI and SpaceX will be fine; their lawyers will sort out the rest. The developer who just wanted their editor to keep working is the one who has to plan around November. That asymmetry, not the feud, is the story.