OpenAI Will Cut Off Cursor's Direct Access to Its Models

What happened

OpenAI has notified SpaceX that it intends to wind down the contract letting Cursor’s code editor draw on OpenAI’s models, with a shutoff date of November 12, 2026 — which OpenAI says is the latest date its contract allows, timed to give developers the maximum possible notice.

OpenAI’s stated reason: “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 points to two prior episodes: after Musk acquired Twitter, Twitter “broke the terms of our contract (alongside many others),” OpenAI says, and Musk “admitted under oath earlier this year that xAI, now also part of SpaceX, had violated OpenAI’s terms of service.” That admission traces to courtroom testimony: during the April 28–May 2, 2026 trial in Musk v. Altman, Musk told OpenAI’s lawyer under questioning that xAI “partly” distills OpenAI’s models to train Grok, a practice OpenAI’s terms of service prohibit. OpenAI also cites a limited contractual window to cancel after a change of control, and “a new level of accountability” tied to its upcoming model, Astra, as part of its reasoning for not extending future models to Cursor. The company adds that it has worked with Cursor for nearly four years and has “enormous respect for their team, their product, and what they’ve built for the developer community.”

The trigger for the decision was a change of ownership. SpaceX closed an all-stock acquisition of Cursor’s parent company, Anysphere, in a deal SpaceX itself valued at roughly $60 billion, according to CNBC — a figure not independently confirmed; the deal was announced June 16, 2026 and closed in mid-August 2026, days before OpenAI’s announcement.

Cursor is not left with a single supplier. Cursor’s own documentation lists models from OpenAI, Anthropic, Google and xAI, plus Cursor’s in-house models, as available inside its editor. According to Cursor co-founder and CEO Michael Truell — the only source for this figure, and one with a stake in minimizing the disruption — OpenAI models account for only about 5% of Cursor’s user traffic.

Rivals moved quickly. Hours after OpenAI’s announcement, Anthropic co-founder Tom Brown posted that “Cursor has been a trusted partner of Anthropic since Sonnet 3.5. We’ll continue to increase compute to support Claude models in Cursor and are excited for what comes next with them at SpaceX.” Musk’s own response, posted publicly, was: “I couldn’t care less. Scam Altman and Greg Stockman are utterly untrustworthy assholes who stole an open source nonprofit.”

The announcement lands against the backdrop of an already public feud: Musk had earlier sued OpenAI and Sam Altman alleging betrayal of the company’s nonprofit mission, and a federal jury ruled against Musk on that suit in May 2026.

What this means (and what it does not)

The immediate consequence is contractual, not technical: on November 12, 2026, Cursor loses the ability to route requests to OpenAI’s models directly through this agreement. OpenAI’s framing — that it cannot trust SpaceX to honor its terms of service, given Twitter’s and xAI’s histories — casts the decision as a compliance and safety measure. That framing also happens to serve OpenAI commercially: cutting off Cursor removes a rival-owned distribution channel and denies a competitor, now under SpaceX/xAI, access to OpenAI’s upcoming Astra model. Anthropic’s public pledge to add compute for Cursor, posted within hours, is a direct beneficiary’s response, and Anthropic separately has its own commercial compute relationship with SpaceX. Musk, the party whose companies’ conduct is cited as the justification, has an obvious incentive to dispute OpenAI’s account, which his public reply reflects.

What does not follow is that Cursor becomes unusable, or even that most of its traffic is disrupted. Cursor’s own documentation shows multiple model providers already available in the editor, and the 5% figure — again, from Truell alone, not independently measured — would if accurate mean the bulk of Cursor’s usage was already running elsewhere before this announcement. Nor does OpenAI’s account establish, beyond its own assertion, that Twitter actually breached its contract, or reveal what the disputed “change of control” clause actually says. Whether this signals a broader OpenAI policy toward other Musk-linked resellers, or was shaped by input from Microsoft or other OpenAI stakeholders, is not addressed in anything reviewed here.

What we still do not know

OpenAI’s announcement does not name which alternative model providers Cursor uses or describe the cost or difficulty of moving away from OpenAI’s models — that context comes only from Cursor’s own documentation and from Truell’s separate statement, not from OpenAI. No source reviewed includes the actual contract language between OpenAI and Cursor, including the specific change-of-control cancellation clause OpenAI cites, so its exact terms are unverified. OpenAI’s claim that Twitter breached its contract “alongside many others” after Musk’s acquisition rests on OpenAI’s own assertion alone; no independent account of that alleged breach was found. Whether other developer tools or AI resellers tied to Musk-controlled entities face similar terminations is unaddressed. The 5% traffic figure has no independent measurement behind it. Whether Microsoft, OpenAI’s largest investor, had any role in the decision is unknown. And whether Cursor or SpaceX intends to contest the cancellation, negotiate an extension, or pursue legal action had not been addressed in any source as of this writing.