OpenAI is ending its contract with the AI coding tool Cursor. The company says the decision followed SpaceX's acquisition of Cursor, and that the reason is that Elon Musk's companies have repeatedly broken contracts.

The contract will terminate on November 12, 2026. OpenAI says that is the maximum notice period the agreement allows; the contract includes a clause giving it a limited window to exit after a change in ownership, and the acquisition triggered it. The company says it can no longer be confident that SpaceX will follow the terms of service.

"It boils down to trust"

OpenAI's Thibault Sottiaux sums up the decision in that sentence. He also goes out of his way to signal that this is a specific move against Musk, not a stance against the broader AI coding tool landscape or its developer community.

The door is not fully shut. Users who reach GPT models through Cursor can still use their own OpenAI API keys, and OpenAI will keep providing access through its own IDE extensions for Cursor.

What changes for developers

  • Until November 12 nothing changes; the contract ends on that date.
  • With your own key you can keep using GPT models inside Cursor.
  • OpenAI's own extensions will keep working for Cursor.
  • The Claude side is unaffected — Anthropic says it will expand capacity.

The record being cited

The suspension is the latest chapter in a years-long feud between Musk and OpenAI that has also played out in court. OpenAI accuses Musk of breaching contracts after his acquisition of Twitter and of trying to extract data.

The episode it points to: OpenAI had a licensing deal with Twitter giving it access to the full tweet feed for about $2 million a year to train ChatGPT. When Musk learned of the agreement in December 2022, he decided the price was too low and cut OpenAI's access.

During the legal battle Musk also admitted using data from other labs' models to train his own Grok models — a practice known as distillation, which likewise violates OpenAI's terms of service.

Anthropic's irony

Anthropic moved to fill the gap immediately. Co-founder and Chief Compute Officer Tom Brown wrote that Cursor has been a trusted partner since Sonnet 3.5 and that the company will keep expanding capacity for Claude models inside Cursor.

There is an irony there. Anthropic has itself pulled API access from partners more than once: it blocked Windsurf after its planned sale to OpenAI, and revoked OpenAI's own access over unauthorised use of Claude for benchmarks.

The "reliable partner" role, in other words, is one nobody in this industry holds for long. Because model providers are simultaneously infrastructure vendors and competitors, which models a tool may offer stops being a technical decision and becomes a commercial one.

For developers the conclusion is plain: which model you can use depends not on your preference but on the relationship between suppliers. Committing to a single tool and a single model means returning to the negotiating table every time those relationships shift.