Cursor’s ‘Composer 2’ model is apparently just Kimi K2.5 with RL fine-tuning. Moonshot AI says they never paid or got permission

A significant controversy has erupted in the AI developer community after Moonshot AI, the Chinese company behind the popular Kimi chatbot, publicly accused code editor startup Cursor of improperly using its technology. According to Moonshot, Cursor's newly touted 'Composer 2' model, which the company integrated into its AI-powered IDE, is not an original creation. Instead, Moonshot claims it is essentially their Kimi K2.5 model that has undergone additional reinforcement learning (RL) fine-tuning. The core of the allegation is that this usage occurred without any formal agreement, financial compensation, or attribution to Moonshot as the original model creator.

The dispute centers on the murky ethics of model reuse in the open-source era. While many AI models are publicly available, their licenses often have specific requirements for attribution, commercial use, or sharing derivatives. Moonshot's statement suggests Cursor may have violated these terms. For developers, this raises critical questions about the provenance of the tools they use daily. If true, it means a paid feature within Cursor's professional editor is built on potentially unlicensed technology, which could have legal and operational ramifications for the company and its users.

This incident is a stark reminder of the growing pains in the fast-moving AI industry. As companies race to integrate cutting-edge capabilities, the lines between inspiration, fine-tuning, and infringement can become blurred. The outcome of this dispute could set an important precedent for how model licensing is enforced and how credit is assigned in the collaborative yet competitive world of AI development.