OpenAI sets November 12 cutoff for Cursor after SpaceX acquisition
- OpenAI told SpaceX it intends to end the contract supplying OpenAI models to Cursor, with a proposed shutoff date of November 12, 2026.
- OpenAI says it is giving Cursor the contract's maximum notice period so developers retain access longer, but will not provide future models to Cursor before the cancellation takes effect.
- The company cites past alleged contract breaches by Musk companies, including Musk's sworn admission that xAI violated OpenAI's terms by distilling OpenAI model data.
- Cursor's custom agreement gives OpenAI a limited cancellation window after a change of control; OpenAI says its upcoming Astra model requires greater accountability for terms-of-service compliance.
- OpenAI says it has worked with Cursor for nearly four years and will support developers affected by the transition, without specifying the support measures.
Hacker News 의견들
Cursor's harness is good, but I rarely used OpenAI models there. Composer and grok-cursor are competitively priced with generous plan limits, and Sonnet got too expensive per token for my use.
Cursor currently lets users bring their own OpenAI token. But without OpenAI access for prompt engineering and evals, I wonder whether Cursor will keep supporting new OpenAI models through BYOK.
This was inevitable given the Altman-Musk feud and Musk's admission of distillation. I'm curious whether Anthropic will act similarly, though its SpaceX hardware dependence may make that harder.
I read the statement as OpenAI keeping current models available until the cutoff while blocking any newly released models before then. The wording about not providing future models seems deliberate.
As a Cursor and Claude subscriber who does not pay OpenAI directly, this will probably push me back toward Anthropic. I did find GPT 5.6 Sol useful in Cursor now and then.
Cursor's resale model always looked temporary. Third-party models are costly there unless you spend thousands of dollars, so its value is mostly in using Grok or Composer.
My alternate take is that open-weight models and model routers are the better value anyway, subsidized plans aside.
Every model provider has an incentive to capture API value itself. Cursor can fall back on its own models, open-weight APIs, or its own hardware, and those models are good enough for many tasks.
This reads like protection for Astra against distillation. I want to know whether Astra launches before or after November 12.
Calling it an excuse misses the stated evidence. Musk admitted under oath that xAI distilled OpenAI models, and I would not keep selling to a competitor that was openly violating my terms.
OpenAI will likely get a bump in Codex users trying new models, and it can point to that growth afterward.
I suspect OpenAI is building a full Cursor alternative around Codex and may announce it in September. Otherwise it risks losing developers who are used to Cursor's workflow.
If providers blacklist Cursor, it may have to move to open-weight models. At that point, I wonder what distinguishes it from OpenRouter beyond the harness.
Cursor already has open models in its pricing documentation. Its popular harness is the meaningful difference.
I doubt this hurts Cursor much. Most users use Auto mode or models like Composer-2 and Grok, and OpenAI's withdrawal matters far more if Claude also leaves.
I have rarely seen one large company accuse another this plainly, without legalistic filtering. It makes the Sam and Elon feud look unusually direct.
I do not see why Cursor users pay extra for OpenAI or Anthropic models unless they value its tooling that much. Codex often scores near the top among harnesses, and Cursor should host more open models.
OpenAI says developers are most affected and promises transition help. That makes me suspect it may be preparing a Cursor-like product of its own.