Thursday, 12 November, is the day OpenAI's models go dark inside Cursor. OpenAI told Cursor's owner, SpaceX, on 28 August that it intended to wind down model supply to the AI coding editor, proposing 12 November as the cutoff — roughly three months' notice. That date has now arrived: Cursor's bundled subscription no longer includes OpenAI models, and the company has said future OpenAI models will not be supplied to it either, effectively ending direct access for as long as SpaceX owns the business.
The rupture traces back to corporate ownership. SpaceX completed its $60 billion all-stock acquisition of Cursor's parent company, Anysphere, on 14 August, in what reports describe as the largest startup buyout of its kind. With Cursor now sitting inside a competitor that also runs its own AI efforts, OpenAI's decision to stop supplying it reads less as a technical dispute than as a supply-chain move by a supplier wary of arming a rival.
For developers the change is narrower than the headlines suggest. Cursor has said OpenAI models accounted for only about 5% of its traffic, and users retain alternatives: Cursor's in-house Composer model, third-party models from Anthropic (whose access is unaffected), and — notably — SpaceX-owned Grok. Some workflows can still reach OpenAI through a bring-your-own-key setup billed directly by usage, though the seamless, bundled experience changes.
The episode has become a talking point about a structural risk in modern software development. AI coding assistants are now embedded in daily engineering work, and this shows that the models powering them can be switched off by a supplier's commercial strategy rather than by any technical failure. Commentary from industry outlets, including CIO.com, frames tools like Claude Code, Codex and Cursor as part of the software supply chain — often adopted by individual developers without the formal review a dependency of that importance would normally attract.
What to watch next is whether Cursor's own models and the Grok integration can absorb the displaced usage without a visible drop in quality, and whether OpenAI's move becomes a template for how frontier labs treat competitors who depend on their models. For teams standardised on OpenAI models inside Cursor, the practical task is unglamorous but immediate: confirm which model each workflow now calls, and re-test prompts that were tuned to a specific OpenAI model.
Sources
- cio.comClaude Code, Codex and Cursor have become part of your software supply chain without approval
- medium.comCursor Users Have 3 Days. Here Are the 2 Things to Check
- tldr.techOpenAI plans to end Cursor model supply
- catdoes.comCursor vs ChatGPT: Which Is Better in 2026?
- testmuai.comCursor model access and cutoff reporting (InfoWorld, Aug. 31 2026)

