OpenAI Is Cutting Off Cursor: What to Do Before the November 12 Shutoff

Update — August 29, 2026: OpenAI has confirmed it is winding down the contract that supplies its models to Cursor, with a proposed shutoff of November 12, 2026. If you run GPT-5.x or Codex-family models in Cursor daily, the practitioner steps are below — start with switching your default model this week.

On August 28, 2026, OpenAI published a post titled "Our decision on Cursor following its acquisition by SpaceX," announced on X the next day. The decision: OpenAI will stop providing its models to Cursor, and it has proposed November 12, 2026 as the shutoff date. OpenAI framed the date as generous — it says it is giving "the maximum notice our contract allows" specifically "to maximize developer access time."

This lands roughly two weeks after SpaceX officially closed its $60 billion acquisition of Cursor, which put Cursor inside the same product org that ships Grok. The Musk–OpenAI feud is the backdrop, but the thing that matters for your week is narrower: a model provider you may depend on inside your editor has set a date to leave.

TL;DR — what Cursor users need to know

Why OpenAI is doing this

OpenAI was direct about the reason. It wrote that it "cannot be confident that SpaceX will use our technology within our terms of service, based on our experience with Elon Musk's companies violating contracts." The post cites two specific incidents: after Musk acquired Twitter — now part of SpaceX — "the company broke the terms of our contract (alongside many others)," and Musk "admitted under oath earlier this year that xAI, now also part of SpaceX, had violated OpenAI's terms of service."

OpenAI said its large-partner contracts include custom terms plus a change-of-control cancellation window, and that it wants "a new level of accountability to ensure our upcoming model, Astra, is being used in accordance with our terms." In other words: the acquisition triggered a clause OpenAI had negotiated for exactly this situation.

OpenAI also went out of its way to praise Cursor — roughly four years of collaboration, and "enormous respect for their team, their product." It said it is "ready to go above and beyond to support" affected developers. This is a corporate-relationship decision, not a judgment on Cursor's engineering.

The community reaction was less measured. Developer threads mixed genuine frustration with snark — jokes about the recent five-hour usage limits and about all the compute that just got "freed up." If you want the history of the underlying dispute, see our Musk vs. Altman feud timeline. Everything past this section is about what you actually do.

Step 1 — Switch your Cursor default model this week

Do not wait until November. The single highest-value move is to stop routing your work through OpenAI models inside Cursor now, so your muscle memory and your prompts are already tuned to the replacement before the cutoff forces it.

• Set an explicit default. In Cursor's model settings, pick a specific Anthropic Claude model or a Google Gemini model as your default rather than leaving requests on Auto or Cursor Router. Router and Auto make per-request provider choices that can still weight toward OpenAI models until access is cut — an explicit default removes that surprise.
• Re-test your common tasks. Run your typical agentic edits, refactors, and debugging sessions on the new default. Claude models are the closest drop-in for Codex-style long-running agentic work; Gemini models are strong on large-context reads and are cheaper per token for high-volume chat.
• Update team defaults. If you administer a Cursor Team or Enterprise plan, change the org default and tell people why. A silent shift on November 12 is a bad way for a team to find out.
• Watch your loops. If you run scripted Cursor loops, update the model parameter in those scripts now and confirm the output quality holds.

Step 2 — Be realistic about bring-your-own-key

The hopeful theory is that a personal OpenAI API key, plugged into Cursor's BYOK setting, keeps GPT-5.x alive after the corporate contract ends. Treat that as unreliable.

The wind-down is a decision about Cursor as a customer. Traffic from Cursor's client — even carrying your own key — still flows through Cursor's integration surface, which is precisely what OpenAI is severing. OpenAI has not published BYOK carve-out terms for this case, and building a migration plan around an untested loophole is how teams get caught flat on cutoff day. If BYOK does keep working, treat it as a bonus, not the plan.

If you want to keep using OpenAI models directly, the durable path is to call them outside Cursor — through the Codex CLI or the Codex-as-a-platform harness, which talk to OpenAI's API without Cursor in the path.

Step 3 — Evaluate a second harness before you need it

This is the moment to learn that your editor and your model provider are separate choices. An agent harness is the loop — context assembly, tool calls, file edits, verification — and it is swappable independently of which model runs underneath.

Our Claude Code vs. Cursor vs. GitHub Copilot comparison breaks down where each one fits. For a provider-neutral option, see the OpenCode guide. Whichever you pick as a backup, actually run a real task through it once — the Claude Code loop guide is a good first exercise — so the fallback is proven, not theoretical.

Step 4 — De-risk single-vendor workflows generally

The specific fix here is a model swap. The general lesson is bigger, and it is worth internalizing while the example is fresh:

• Any critical workflow on a single vendor relationship can be severed by events you don't control. Here it was an acquisition triggering a contract clause. It could just as easily be a price change, a rate-limit cut, a policy shift, or a deprecation.
• Keep the harness and the model loosely coupled. Know how to run your core tasks in at least two harnesses, and default to a model provider whose access does not depend on who owns your editor.
• Don't hard-code one model everywhere. In scripts, CI, and internal tools, put the model name behind a single config value so switching providers is a one-line change, not a codebase sweep.
• Notice the structural irony. Anthropic trains partly on SpaceX's Colossus supercomputer while SpaceX now owns Cursor — infrastructure partner and product rival under one roof. Alignments between AI companies shift fast; your workflow should not assume today's arrangement is permanent.

What to watch between now and November 12

• Whether the date moves. OpenAI called November 12 a proposed shutoff. A negotiated extension — or an earlier cut if the relationship sours — is possible.
• A Cursor response. As of the announcement, Cursor had not published its own statement on migration support, credits, or default-model changes for affected users.
• BYOK clarification. If OpenAI or Cursor publishes explicit terms for personal-key usage after the contract ends, that changes Step 2.
• Router and Auto behavior. If Cursor rebalances Router away from OpenAI models ahead of the cutoff, your requests may shift before November 12 without any action from you.

Related on explainx.ai

• SpaceX closes its $60B Cursor acquisition — Cursor joins the Grok team
• SpaceX is acquiring Cursor for $60 billion — the SEC 8-K breakdown
• Claude Code vs. Cursor vs. GitHub Copilot: the full comparison
• What is an agent harness? The complete guide
• Cursor Router — automatic model selection across Claude, GPT, and Grok
• OpenCode — the open-source, provider-neutral AI coding agent
• How to run loops in Cursor · Claude Code · Codex CLI
• Codex as a platform — OpenAI's open agent harness
• Musk vs. Altman feud — the history behind this decision
• Anthropic secures SpaceX Colossus 1 supercomputer

Primary source: OpenAI, "Our decision on Cursor following its acquisition by SpaceX," openai.com, August 28, 2026.

Facts and dates are accurate as of August 29, 2026. November 12, 2026 is a proposed shutoff date and could move; no Cursor migration-support details had been published at the time of writing. Verify current model availability in Cursor's settings before changing team defaults. Follow @explainx_ai for updates.