OpenAI Cuts Off Cursor’s Model Access After SpaceX Acquisition

On this page
OpenAI is cutting off Cursor’s direct access to its models on November 12, 2026, roughly two weeks after Elon Musk’s SpaceX closed its $60 billion purchase of Cursor’s parent company, Anysphere. OpenAI says it triggered a change-of-control clause in its supply contract because it cannot trust that SpaceX will keep to its terms of service, pointing to past contract disputes with Musk-owned companies. Cursor says OpenAI models handle only about 5% of its traffic and that talks are ongoing. I have been watching the model-supply business for a while, and this is the clearest sign yet that access to frontier models has become a lever companies will pull against each other.
What OpenAI actually announced
Late on Friday, August 28, OpenAI told SpaceX it intends to wind down the contract that provides OpenAI models to Cursor, the AI coding tool. The company set a “proposed shutoff date” of November 12, 2026, and said it is giving “the maximum notice provided by our contract.” It also said it will not ship any future models to Cursor during the wind-down period.
The mechanism matters here. OpenAI signs custom agreements with large partners rather than letting them run on the standard API terms, and those custom agreements include a provision that lets OpenAI cancel within a limited window after the partner changes hands. SpaceX completed its acquisition of Anysphere on August 14, according to financial filings, which opened that window. OpenAI chose to use it.
| Detail | Where things stand |
|---|---|
| Announcement | Late Friday, August 28, 2026 |
| Proposed cutoff for OpenAI models in Cursor | November 12, 2026 |
| Trigger | Change-of-control clause in OpenAI’s custom contract |
| Corporate event | SpaceX closed its $60B buy of Anysphere on August 14, 2026 |
| OpenAI’s stated reason | Cannot be confident SpaceX will follow OpenAI’s terms of service |
| Future models (including “Astra”) | Will not be provided to Cursor |
| Share of Cursor traffic on OpenAI models | About 5%, per Cursor CEO Michael Truell |
| Filling the gap | Anthropic reportedly adding compute for Claude models in Cursor |
The reason OpenAI gave
OpenAI’s explanation is unusually blunt for a corporate statement. “We are making this 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,” the company wrote. It cited two specific episodes: after Musk acquired Twitter (now folded into SpaceX along with xAI), the company broke the terms of an OpenAI contract, and under oath earlier this year Musk acknowledged that xAI had violated OpenAI’s terms of service, which are similar to xAI’s own.
There is a second thread in the statement about model safety. OpenAI framed its decision partly around “a new level of accountability” for its upcoming model, referred to as Astra, and said it wants to be sure that model is used in line with its terms. OpenAI paused reinforcement-learning work on Astra in early August after the model hit an internal cybersecurity threshold the company classifies as critical, so the caution is consistent with how OpenAI has been talking about that system all month. Withholding future models from a partner it no longer trusts to enforce usage rules follows from that posture.
What Cursor and Musk said back
Cursor co-founder and CEO Michael Truell responded on X, the SpaceX-owned network. “OpenAI models serve about 5% of Cursor user traffic, and we’re speaking with the OpenAI team to resolve this,” he wrote. “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.” That last phrase is the crux of the whole situation, and I will come back to it.
Musk was less measured. In a post on X the next day he said he “couldn’t care less,” called OpenAI’s Sam Altman and Greg Brockman untrustworthy in cruder terms, and repeated his long-running claim that they “stole an open source nonprofit.” Musk co-founded OpenAI as a nonprofit in 2015, left the board in 2018, and sued the company in 2024 over its shift to a for-profit structure. He lost that case earlier this year and has said he will appeal. None of that backstory is new, but it explains why a routine-sounding contract decision reads like another round in a personal feud.
Who picks up the slack
Cursor is a multi-model tool. It routes requests to models from OpenAI, Anthropic, Google, and, since the SpaceX deal, Musk’s own Grok. Losing OpenAI removes one option from the menu rather than breaking the product, which is why Truell can call it 5% of traffic and largely mean it. Developers who specifically prefer GPT-series models for certain tasks will feel the change more than the average number suggests, and anyone who built a workflow around a particular model will have to test alternatives before November.
Anthropic appears ready to absorb most of that demand. Reporting around the announcement says Anthropic is adding compute capacity to support Claude models inside Cursor. That tracks with where enterprise coding spend has been moving this year, and it means the practical outcome for many Cursor users will be a nudge toward Claude rather than a loss of capability. If you want the wider context on how fast this category is moving, our earlier piece on Meta’s Muse Code coding agent covers how many players are now fighting over the same developers.
Why this is bigger than one integration
For years the working assumption in AI tooling was that model providers wanted their models everywhere. Distribution was the goal, and a tool like Cursor was a channel worth cultivating. This decision breaks that assumption. When the company on the other side of the integration is owned by a direct competitor, model access stops being a neutral utility and becomes something a provider can grant or withdraw for strategic reasons.
That has a cost for everyone building on top of third-party models. “Neutral infrastructure,” to use Truell’s words, is only neutral until ownership changes or a rivalry turns sharp. A tool that depends on several providers is exposed to all of their corporate politics at once. The lesson for developers is not to panic about Cursor specifically, but to treat single-model dependencies as a risk to plan around, keep prompts and evaluations portable, and assume that any given model could become unavailable on short notice. SpaceX’s own $60 billion bet on Cursor, which we covered when the acquisition was announced, was always going to invite exactly this kind of response from competitors.
What to watch next
Three things will tell us how this plays out. First, whether OpenAI and Cursor reach any kind of settlement before November 12, which Truell says they are discussing. Second, whether other model providers with Cursor integrations, particularly Google, take a similar view now that a competitor owns the platform. Third, whether SpaceX leans harder on Grok inside Cursor to reduce its reliance on outside models altogether, which would be the logical endpoint of owning the tool. The vibe-coding market is large enough now that these decisions carry real revenue, and OpenAI is heading toward a public offering, so expect every move to be read through that lens.
Frequently asked questions
Will Cursor stop working in November?
No. Cursor will keep running with models from Anthropic, Google, and Grok. Only OpenAI’s models are being removed, and Cursor says they account for about 5% of usage. You will need to switch to another model for tasks where you currently pick a GPT-series option.
Can I still use OpenAI models in Cursor with my own API key?
OpenAI’s statement targets the supply contract between OpenAI and Cursor rather than individual users, but the company has said it will not provide its models to Cursor as it winds the agreement down. Personal API-key workarounds may not cover every feature, and OpenAI could restrict them further, so do not count on that path as a long-term fix.
Why can OpenAI cancel the contract at all?
Because its custom agreement with Cursor includes a change-of-control clause that allows cancellation within a limited window after the company is sold. SpaceX closing its purchase of Anysphere on August 14 started that clock, and OpenAI acted inside the window.
Is this about safety or about competition?
OpenAI gives both reasons: distrust that SpaceX will enforce its terms of service, and accountability for its upcoming Astra model. The timing, right after a direct competitor bought the platform, means the competitive read is hard to separate from the stated safety concerns.
Source: CNBC, plus OpenAI’s public statement and posts on X from Michael Truell and Elon Musk.
