← All posts / Industry

OpenAI Cuts Off Cursor: The November 12 Deadline That Ends Neutral AI Coding Tools

OpenAI invoked a change-of-control clause to terminate its Cursor contract after SpaceX's $60 billion acquisition, setting a November 12 model shutoff and declaring that model supply is no longer neutral infrastructure.

OpenAI Cuts Off Cursor: The November 12 Deadline That Ends Neutral AI Coding Tools

On Friday night, August 28, OpenAI notified SpaceX that it intends to terminate the contract that supplies OpenAI models to Cursor, the AI coding assistant now owned by Elon Musk’s space company. The proposed shutoff date is November 12, 2026 — the maximum notice period allowed under a change-of-control clause in the agreement, which gave OpenAI a limited window to exit after Cursor’s maker Anysphere was acquired.

The move is the most consequential fallout yet from SpaceX’s record $60 billion all-stock acquisition of Anysphere, which closed on August 14 after a two-month regulatory review. Cursor, the editor that helped ignite the “vibe coding” movement, had been one of OpenAI’s most prominent distribution partners — and one of the very first commercial users of its API. Now, roughly ten weeks after the acquisition became effective, that decade-defining partnership is being unwound by its supplier.

What OpenAI actually said

The stated rationale is trust. “It boils down to trust,” OpenAI’s Thibault Sottiaux said, explaining that the company cannot be confident SpaceX will follow the terms of service once Cursor sits inside Musk’s corporate empire.

The decision, announced Friday night, leans on two arguments:

  1. A track record of broken contracts. OpenAI accuses Musk of breaching agreements after his acquisition of Twitter. The AI lab once had a licensing deal with Twitter that gave it access to the full tweet firehose for roughly $2 million a year to train ChatGPT; when Musk discovered the arrangement in December 2022, he considered the price far too low and cut off access. Musk has also admitted, during the pair’s ongoing legal battles, to using data from other labs’ models to train his own Grok models — the practice known as distillation, which violates OpenAI’s terms of service.
  2. Safety accountability. With frontier capabilities climbing, OpenAI frames the cutoff as a matter of “a new level of accountability” for partners handling its models — a pointed argument given reports that its upcoming Astra model carries advanced cyber capabilities.

Notably, Sottiaux went out of his way to signal that this is a specific move against Musk, not against the AI coding tool landscape or its developer community at large. Users who bring their own OpenAI API keys to Cursor will still be able to use them, and OpenAI says its own IDE extensions for Cursor continue to work.

Cursor’s response: five percent

Cursor co-founder Michael Truell responded quickly, saying OpenAI models account for only about five percent of Cursor’s AI traffic, and that the company is in talks with OpenAI to find a solution.

“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,” Truell wrote.

That last phrase is the one that will haunt the industry. Cursor built a multi-billion-dollar business on the assumption that foundation models are neutral infrastructure — commodity intelligence that any tool vendor could rent on equal terms. OpenAI has now demonstrated, in a single contract clause, that the assumption does not hold when the customer becomes a competitor’s subsidiary. Model supply is a strategic asset, and it will be allocated strategically.

The precedent cuts both ways

OpenAI knows how being cut off feels, because it has been on the receiving end twice. In June 2025, Anthropic blocked the coding tool Windsurf from accessing Claude models after reports surfaced that OpenAI wanted to acquire Windsurf. Two months later, in August 2025, Anthropic revoked OpenAI’s own API access on the grounds that OpenAI teams had apparently used Claude for internal benchmarks ahead of the GPT-5 launch.

Each of those incidents was read at the time as an escalation. Together with Friday’s decision, they sketch a new normal: as coding assistants consolidate into the hands of a few rival camps — OpenAI, Anthropic, Google, and now SpaceX/xAI — model access is becoming a battlefield rather than a marketplace. The Wall Street-era assumption that API access survives corporate warfare is dead.

Why the timing matters

The November 12 date lands at a loaded moment. SpaceX closed the Anysphere deal on August 14; OpenAI moved within two weeks, using the narrow cancellation window the change-of-control clause provided. Waiting longer would have meant normalizing the new ownership and weakening the contractual exit. The swiftness is itself the message: OpenAI treated the acquisition not as a business development event but as a structural threat.

For developers, the practical impact is smaller than the headlines suggest — Cursor’s traffic is dominated by other models, users with their own API keys keep access, and OpenAI’s extensions continue to function. But for tool builders, the lesson is structural: if your product depends on a rival’s models, your moat is rented. Expect the next generation of coding tools to hedge aggressively across model vendors, prioritize open-weight alternatives, and negotiate explicit neutrality guarantees — or accept that neutrality was never really on offer.