OpenAI Cuts Cursor's Access to GPT Models Over SpaceX Deal

If you write code inside Cursor and lean on GPT-5 or GPT-6 for autocomplete and chat, this week is when that habit started breaking. OpenAI has begun winding down the contract that let Cursor offer its models directly, and the cutoff date the two companies settled on was November 12 — yesterday.

Why OpenAI is walking away

The root of this isn't a technical falling-out. It's ownership. Cursor, the AI-powered code editor that millions of developers use daily, was acquired by SpaceX earlier this year. OpenAI's contract with Cursor included a change-of-control clause — a standard legal term that lets either side cancel a deal if the other company gets bought by someone they don't want to be in business with. OpenAI decided to use it.

In its own announcement, OpenAI didn't dress this up as a dispute over pricing or performance. It pointed straight at who now owns Cursor:

"We cannot be confident that SpaceX will use our technology within our terms of service."

SpaceX, of course, is Elon Musk's company, and Musk also runs xAI — a direct OpenAI competitor. OpenAI called the call "incredibly tough" and said it would hold off the cancellation as long as its contract allowed rather than cutting access overnight.

What actually stops working

This isn't a full blackout. Cursor's CEO has said OpenAI's models account for only around 5% of all traffic on the platform, since most developers already default to Anthropic's Claude, Google's Gemini, or Cursor's own in-house model called Composer. Here's the practical breakdown:

• Built-in, one-click access to GPT models inside Cursor goes away.
• Developers can still use GPT models through BYOK — "bring your own key" — which means plugging in your own personal OpenAI account and paying for it directly, instead of it being bundled into Cursor.
• OpenAI's own coding extensions for other editors keep working as before.
• Claude, Gemini, and Grok-based models inside Cursor are unaffected.
• Future OpenAI models won't be added to Cursor at all, BYOK or not, going forward.

For a solo developer tinkering on side projects, that's an annoyance. For a company that built CI pipelines, automated code-review bots, or internal tooling around a specific GPT model inside Cursor, it's a dependency they now have to rebuild on someone else's terms.

Why this should worry Indian dev teams specifically

India has quietly become one of the biggest markets for AI coding assistants — GitHub's own data has repeatedly shown Indian developers among the fastest-growing user bases for tools like Copilot and Cursor, and most mid-size Indian startups and IT services firms have folded some AI-assisted coding into their workflow over the past two years. None of those teams had any say in a boardroom dispute between two American companies, yet they're the ones who'll spend a weekend migrating prompts, re-testing outputs, and rewriting internal docs because of it.

It's a sharp reminder of a risk that doesn't get discussed enough locally: building a product or a workflow around a single AI vendor's model is a business decision, not just a technical one. If your team's code reviews, test generation, or documentation pipeline leans on one model from one company, you're exposed to decisions made entirely outside your control — an acquisition, a policy change, a pricing shift. Indian engineering leads evaluating AI tooling contracts would do well to ask the same question OpenAI effectively just asked Cursor: what happens if ownership changes?

This also comes in the same month Cursor users were targeted by a tampered extension that planted an SSH backdoor on the platform — a separate incident, but together they make the case that teams shouldn't treat any single AI coding tool as a fixed, risk-free part of their stack.

The bigger pattern here

What's really happening is that the AI coding tool market is starting to fracture along ownership lines rather than just technical merit. When the company that makes your editor also owns the rocket company that owns the EV company that owns a rival AI lab, "which model should I use" stops being a pure engineering question and starts being a corporate-politics one. Expect more of this, not less, as AI labs and the companies that build on top of them keep getting bought, merged, and reshuffled — and expect the developers in the middle, wherever they are, to be the ones who have to adapt fastest.