OpenAI Ends Cursor Model Access After SpaceX Acquisition
OpenAI says it has told SpaceX that it will wind down its contract providing OpenAI models to Cursor, with a proposed shutoff date of November 12, 2026. OpenAI points to a change-of-control clause and concerns about compliance after SpaceX acquired Cursor's parent for $60 billion. Cursor says only about 5% of its traffic uses OpenAI models.
On this page
OpenAI has proposed a November shutoff
OpenAI announced on Friday that it had notified SpaceX of its intention to wind down the contract that supplies OpenAI models to Cursor. The company proposed November 12, 2026 as the shutoff date and said it would give the maximum notice allowed by the contract. It also said it would stop supplying future models during that transition [1].
The announcement is explicit rather than speculative. OpenAI framed the move as a contract decision, not a product review, and identified SpaceX's acquisition of Anysphere, Cursor's parent company, as the trigger. CNBC reports, based on financial filings, that SpaceX completed the $60 billion acquisition on August 14 [2].
Cursor says the affected share is small
Cursor CEO Michael Truell wrote on X that OpenAI models served about 5% of Cursor user traffic. He said Cursor was speaking to OpenAI to resolve the issue, described OpenAI's platform as neutral infrastructure that Cursor had trusted for years, and noted Cursor's long history with the company [2].
That percentage may limit the immediate blast radius, but it does not make the transition trivial. The 5% figure does not reveal which workflows, accounts, pricing tiers, or team policies depend on those models. Developers cannot tell from it whether their own completions, refactors, agents, or enterprise rules will change on November 12.
OpenAI's explanation centres on control and compliance
OpenAI said it could not be confident that SpaceX would use its technology within its terms of service, citing its experience with Elon Musk's companies violating contracts [1]. It pointed to Twitter, now part of SpaceX, allegedly breaking an OpenAI contract after Musk acquired it, and said Musk had admitted under oath earlier in 2026 that xAI, also now part of SpaceX, had violated OpenAI's terms [1].
The company also explained the timing. Its custom Cursor agreement had a limited window to cancel after a change of control. OpenAI said it was using the latest cancellation date available while withholding future models, including Astra, from Cursor [1]. Reuters likewise reports that OpenAI custom contracts for large partners are meant to ensure terms compliance and safety at scale, and that the change-of-control clause shaped the notice period [3].
Anthropic is competing for Cursor's usage
Anthropic co-founder Tom Brown said Cursor had been a trusted Anthropic partner since Sonnet 3.5 and that Anthropic would increase compute to support Claude models in Cursor. CNBC notes that Anthropic is also partnering with SpaceX and renting compute capacity from it [2].
The response landed quickly, and some executives criticised it. Replit CEO Amjad Masad reminded Brown of Anthropic's earlier cutoff of Windsurf, while Docker executive Mat Velloso suggested Anthropic should have stayed quiet [2]. The exchange shows how unstable model availability can become when supplier relationships, ownership, and competitive pressure collide.
Model access is now a continuity decision
The practical lesson is not that Cursor is unusable or that every developer should abandon hosted tools. Many will continue using Cursor with Anthropic, other providers, or local models. But the episode shows why model access should be treated as configurable infrastructure rather than a fixed property of an application. A product can keep working while a particular model disappears from it.
Teams using Cursor should identify which features use OpenAI, test an alternative before the proposed deadline, and avoid hard-coding one provider into critical automations. The same review should cover data retention, pricing, rate limits, and fallback behaviour. For individuals, a local fallback does not replace a frontier cloud workflow, but it can preserve ordinary assistance when a hosted provider changes access, terms, or ownership.
Recent disputes now form a pattern. Anthropic's June government-directed model halt fed a European debate about AI sovereignty, and Cursor's OpenAI cutoff arrived after an acquisition [4]. In both cases, users did not lose a feature because it stopped working. They lost access because a provider's obligations and commercial relationships changed.