OpenAI Cuts Off AI Models to Cursor Citing SpaceX Acquisition and Terms of Service Violations
OpenAI is terminating its AI model supply to the coding assistant Cursor, citing concerns that Elon Musk's companies, including SpaceX and xAI, have violated its terms of service by using its technology to train competing systems.
OpenAI has announced it will cease supplying its AI models to the popular coding assistant Cursor, a decision directly linked to SpaceX's recent acquisition of Cursor's parent company, Anysphere. The AI research giant has notified SpaceX that it intends to wind down the contract that provides these models, with a proposed shutoff date of November 12, 2026, the maximum notice period allowed under their custom agreement. OpenAI stated that while they want developers to retain access for as long as the contract permits, they cannot be confident that SpaceX will adhere to their terms of service, citing past experiences with Elon Musk's companies.
This move is characterized by OpenAI as a "change-of-control" decision, rather than an accusation that Cursor itself has violated any rules. OpenAI emphasizes its reliance on custom contracts with large partners to enforce terms of service and maintain safety at scale. The acquisition of Anysphere by SpaceX triggered a clause in the agreement that allowed for cancellation, which OpenAI invoked approximately two weeks after the deal closed. Consequently, future model supply, including the advanced Astra model, has been frozen, while current models will remain operational until mid-November.
Cursor will continue to offer AI models from other providers, including Anthropic, Google, and SpaceXAI, meaning the November 12 cutoff specifically impacts OpenAI's direct in-editor integration. The core of OpenAI's concern stems from alleged past violations by other Musk-owned entities. OpenAI claims that after Elon Musk acquired Twitter, the company breached its contract by cutting off OpenAI's paid license to Twitter data. Furthermore, OpenAI stated that Musk admitted under oath that xAI had violated OpenAI's terms of service by using its outputs to train competing systems.
Given that Cursor will now fall under the same corporate umbrella as xAI, OpenAI views the continued supply of its powerful AI models as too risky. This decision is particularly significant concerning the Astra model, which OpenAI disclosed on August 7 has demonstrated substantial gains in agentic coding and cybersecurity capabilities. OpenAI's internal evaluations indicated that Astra could potentially achieve "Critical cyber capability" under its Preparedness Framework, a threshold that includes autonomously finding and developing zero-day exploits and executing complex end-to-end attacks.
Withholding a model possessing such advanced capabilities from a competitor's platform is a strategic access-control decision with significant commercial implications. OpenAI has expressed respect for the Cursor team, with whom they have collaborated for nearly four years, and pledged to support developers through this transition. Until the November deadline, existing OpenAI models should continue to function within Cursor.
Post-November 12, teams that have standardized on GPT models within the Cursor editor will need to explore alternative solutions. This could involve switching to a different AI model, adopting another integrated development environment (IDE), or potentially using their own OpenAI API key if Cursor permits such a configuration. Enterprises that have treated Cursor as an official OpenAI channel will need to conduct a thorough review of their logging, data handling, and policy adherence within the next ten weeks to ensure continued compliance and security.