共有:

AI Coding Tools

OpenAI cuts Cursor off.

Last month, picking a GPT model inside Cursor was still completely routine. Two weeks after SpaceX's acquisition closed, OpenAI announced it will end model access outright.

AI Navigate Editorial2026.09.207 min read

OpenAI Cursor Owned by SpaceX CUT NOV 12 Before: GPT models freely selectable
FIG. Following the SpaceX acquisition, OpenAI ends model access to Cursor on November 12
01
The Announcement

A decision made two weeks after the deal closed

The trigger was SpaceX completing its acquisition of Cursor's maker, Anysphere.

Anysphere, the company behind Cursor, saw its acquisition by SpaceX close on August 14, 2026. The all-stock deal was valued at $60 billion, with roughly 391 million shares of SpaceX Class A stock issued — reported as one of the largest startup acquisitions on record. Cursor became part of a newly formed division called "SpaceXAI."

Two weeks later, on August 29, OpenAI published a blog post titled "Our decision on Cursor following its acquisition by SpaceX," announcing it will end GPT model access for Cursor effective November 12. Per OpenAI's post, the company says it is giving the maximum notice period allowed under its contract. In parallel, OpenAI said on its own X account that "we know that the people most affected by this decision are the developers who rely on OpenAI models in Cursor," and pledged extra transition support.

02
Why

The reason: contract trust, not technology

OpenAI's stated rationale wasn't technical — it was distrust over whether Musk-linked companies honor their contracts.

$60B
SpaceX's Cursor acquisition, closed 2026.08.14
11.12
OpenAI model access end date
~5%
share of Cursor traffic on OpenAI models

OpenAI's official reason is that it "cannot be confident that SpaceX will use our technology within our terms of service." Per CNBC's reporting, OpenAI points to a history of contract violations by Elon Musk–linked companies, including X (formerly Twitter) and xAI. The same report notes that, during sworn testimony in a federal trial in Oakland, California, Musk himself acknowledged xAI used "distillation" techniques on OpenAI's models to train Grok. That is framed as inconsistent with xAI's own terms banning the same kind of behavior.

Cursor sees it differently. CEO Michael Truell wrote on his own X account that "OpenAI models serve about 5% of Cursor user traffic," adding that Cursor is "speaking with the OpenAI team to resolve this." He also noted that "Cursor was one of the very first users of OpenAI" and that the two companies have worked closely together for years.

Until last monthFrom November 12
GPT models selectable inside CursorDirect OpenAI model access cut off
OpenAI models ~5% of total trafficClaude, Gemini and other models unaffected

The real fight isn't technical.
It's whether Musk's companies keep their contracts.


03
Who It Hits

Who feels this, and who doesn't

Cursor-wide, the impact is contained — but individual workflows built around GPT get a hard deadline.

Engineers

Teams that leaned on GPT-5-series models for coding need to move custom modes and team-wide defaults to other models before November 12. The 5% figure is Cursor-wide — check your own team's actual usage first, since reliance varies a lot team to team.

Leadership & PMs

This doesn't threaten Cursor's viability directly, but it's a clean, shareable case study in the risk of single-vendor model dependence. Expect multi-model tool selection and contract terms to get more scrutiny going forward, well beyond this one incident.

04
What's Next

The next three months

In the short term, expect usage to shift toward Claude, Gemini and other non-OpenAI models. As CNBC also reported, Anthropic has signaled it will add Claude compute capacity for Cursor around this window, positioning rivals to pick up the freed-up demand. Cursor is expected to firm up its own migration support before November 12, though the specifics weren't public at the time of writing. Teams with GPT-specific custom rules or prompts are better off starting re-validation against other models now rather than waiting for the deadline.