OpenAI Cuts Off Cursor's Model Access After SpaceX's $60B Acquisition of Anysphere

OpenAI's move against Cursor isn't a response to anything Cursor did. It's the exercise of a change-of-control clause buried in OpenAI's custom agreement with the company - a provision that gives OpenAI a limited window to walk away from the deal whenever Cursor's ownership changes hands ^[7]. SpaceX's $60 billion all-stock acquisition of Anysphere, Cursor's parent company, closed on August 14, 2026 ^[6]. Two weeks later, on August 28, OpenAI notified SpaceX it intended to wind the contract down, proposing November 12 as the cutoff date ^[1].

OpenAI has been careful to frame this as the maximum notice the contract allows, not a snap decision - the company says it is giving developers as much transition time as the agreement permits, and it still describes Cursor's team and product with genuine respect after what it calls nearly four years of partnership ^[4]. That framing matters: it lets OpenAI draw a sharp line between punishing Cursor, which it insists it isn't doing, and refusing to extend trust to Cursor's new owner, which it is. The mechanism is the real story here - a routine M&A event, not a product failure or a broken agreement on Cursor's side, is what triggered the sanction.

OpenAI's stated rationale isn't about anything Cursor did - it's about not trusting entities connected to Elon Musk to honor contract terms. OpenAI points to Musk's own history: after acquiring Twitter in December 2022, Musk cut off OpenAI's data access despite an existing licensing arrangement worth roughly $2 million a year, and OpenAI has separately alleged that xAI violated its terms of service by distilling OpenAI's outputs ^[2].

OpenAI isn't the first model provider to cut off a coding platform, either. In June 2025, Anthropic blocked AI coding tool Windsurf from Claude access, an earlier precedent of a model provider severing a coding platform's access outright ^[2]. Two months later, Anthropic went further and revoked OpenAI's own API access to Claude entirely ^[2]. Now Anthropic is positioned as Cursor's rescuer: co-founder Tom Brown publicly reaffirmed that 'Cursor has been a trusted partner of Anthropic since Sonnet 3.5' and pledged more compute to absorb the traffic OpenAI is walking away from ^[3]. Read together, these episodes show AI labs increasingly treating model access as a lever to withhold from parties they see as competitors, security risks, or simply untrustworthy - Cursor's situation is the newest entry in that pattern, not a one-off.

By Cursor's own numbers, the immediate damage is small. CEO Michael Truell says OpenAI's models account for roughly 5% of Cursor's user traffic, with the other 95% already running on Anthropic, Google, and xAI models ^[3]. Cursor also has its own proprietary Composer 2.5 model, which reportedly scored 73.7 on SWE-bench Multilingual versus Claude Opus 4.6's 65.9, and is priced at $0.50 per million input tokens - giving Cursor a credible in-house alternative even as OpenAI access winds down ^[3].

Analysts covering the story argue the small blast radius is beside the point. The real lesson for any enterprise building on a single AI vendor is that model access is not a permanent utility - it's a relationship that can be terminated by a contract clause triggered by events completely outside the customer's control, like who buys their vendor's parent company ^[5]. Enterprises still using OpenAI models inside Cursor have roughly 10 weeks from OpenAI's notice to migrate before the proposed shutoff, and switching isn't just a config change - it can require retraining developers, rebuilding integrations and agent configurations, and repeating security and governance reviews ^[5].

OpenAI's public rationale rests on Elon Musk's track record, not abstract policy. The company points to Musk cutting off OpenAI's data access after buying Twitter in December 2022, breaching an existing licensing deal, and to xAI's alleged distillation of OpenAI's outputs in violation of its terms of service ^[2]. OpenAI's Thibault Sottiaux summarized the position bluntly: 'It boils down to trust' ^[2].

But the decision goes beyond severing what Cursor already has - OpenAI is also withholding future models, including one referred to internally as Astra, from ever reaching Cursor under SpaceX ownership ^[1]. That forward-looking piece matters: it isn't just about cleaning up an existing contract, it's about keeping OpenAI's newest, most capable models away from a company Musk now owns, given the xAI distillation allegation OpenAI has already made ^[2]. Musk, for his part, waved the whole decision off: he said he 'couldn't care less' and called Sam Altman and Greg Brockman 'untrustworthy,' reviving his long-running claim that OpenAI's leadership improperly took control of what began as a nonprofit ^[8].