Cursor’s Composer 2 Sparks Controversy Over Undisclosed Chinese AI Model Origins.

AI coding company Cursor found itself at the center of a significant industry discussion this week following the launch of its new model, Composer 2, which it boldly promoted as offering "frontier-level coding intelligence." The announcement, intended to highlight a major step forward in automated coding assistance, quickly took an unexpected turn when an astute observer on X, posting under the pseudonym Fynn, made a startling claim that brought the model’s true origins into question.

Fynn asserted that Composer 2 was, in essence, "just Kimi 2.5" augmented with additional reinforcement learning. Kimi 2.5 is an open-source model that had been recently released by Moonshot AI, a prominent Chinese technology company backed by influential investors such as Alibaba and HongShan (formerly Sequoia China). The accusation immediately raised eyebrows across the tech community, given Cursor’s high-profile status and the broader geopolitical context of the AI industry.

As compelling evidence for the claim, Fynn pointed directly to internal code that seemingly identified Kimi as the foundational model within Cursor’s new offering. The discovery was met with a sarcastic retort from Fynn, who scoffed, "[A]t least rename the model ID," implying a lack of effort to obscure the model’s lineage. This direct evidence suggested that Cursor had not developed its latest innovation entirely from scratch, prompting a wave of questions regarding transparency and the nature of AI model development in a competitive landscape.

The revelation was particularly surprising because Cursor is a well-funded U.S. startup, having achieved remarkable success in a relatively short period. The company had notably raised a substantial $2.3 billion funding round just last fall, pushing its valuation to an impressive $29.3 billion. Furthermore, Cursor has reportedly surpassed $2 billion in annualized revenue, indicating a strong market presence and significant resources. The fact that a company of this stature, with such considerable backing and a reputation for innovation, might rely heavily on an undisclosed open-source base, especially one originating from a Chinese firm, was unexpected. Crucially, Cursor’s initial announcement of Composer 2 made no mention whatsoever of Moonshot AI or its Kimi model, further fueling the controversy.

In response to the mounting pressure and public scrutiny, Lee Robinson, Cursor’s vice president of developer education, soon acknowledged the situation. Taking to X, Robinson confirmed, "Yep, Composer 2 started from an open-source base!" This statement validated Fynn’s initial claim but also provided crucial context. Robinson elaborated that while Composer 2 did indeed leverage an existing open-source foundation, "Only ~1/4 of the compute spent on the final model came from the base, the rest is from our training." He emphasized that the extensive additional training undertaken by Cursor significantly transformed the model, asserting that Composer 2’s performance on various benchmarks is "very different" from that of Kimi’s original version. This suggests a process of fine-tuning, further pre-training on proprietary datasets, and extensive reinforcement learning from human feedback (RLHF) to tailor the model to Cursor’s specific applications and improve its capabilities beyond the base.

Robinson also moved to address concerns regarding the legality and ethical implications of using an open-source model without explicit attribution. He insisted that Cursor’s use of Kimi was entirely consistent with the terms of its underlying license. This point was subsequently corroborated by the official Kimi account on X, which posted a message congratulating Cursor. In its statement, the Kimi account clarified that Cursor utilized Kimi "as part of an authorized commercial partnership" with Fireworks AI. Fireworks AI, a platform known for serving open-source LLMs, likely facilitated the licensing and integration of Kimi-k2.5, ensuring that Cursor’s use was legitimate and compliant with the original model’s terms.

The Kimi account further expressed pride in their model’s contribution, stating, "We are proud to see Kimi-k2.5 provide the foundation." They added, "Seeing our model integrated effectively through Cursor’s continued pretraining & high-compute RL training is the open model ecosystem we love to support." This endorsement from Moonshot AI through its Kimi brand helped to diffuse some of the tension, confirming a legitimate, albeit undisclosed, collaboration rather than an unauthorized appropriation of intellectual property. The sentiment highlighted a common practice in the open-source AI community, where foundational models are often built upon, refined, and specialized by other entities.

However, the question remained: why did Cursor not acknowledge the Kimi base upfront in its initial announcement? Beyond any potential embarrassment associated with not having developed a foundational model entirely from scratch, the decision to omit this detail likely stems from the increasingly sensitive geopolitical climate surrounding artificial intelligence. Building upon a Chinese model might feel particularly fraught right now, especially given the pervasive narrative of an AI "arms race" that is frequently framed as an existential battle for technological supremacy between the United States and China. This rivalry encompasses concerns over national security, economic dominance, data privacy, and the control of critical AI infrastructure.

The industry has previously witnessed the impact of this geopolitical tension. For example, Silicon Valley experienced what was described as "apparent panic" after Chinese company DeepSeek released a highly competitive AI model early last year. DeepSeek’s model challenged the perceived lead of Western AI firms in certain benchmarks and capabilities, underscoring the fierce competition and nationalistic undertones within the global AI development landscape. In this environment, an American company, particularly one with significant funding and market presence, might have strategically chosen to downplay or omit its reliance on a Chinese-developed base model to avoid potential backlash or scrutiny.

Acknowledging the oversight, Cursor co-founder Aman Sanger publicly addressed the issue, stating, "It was a miss to not mention the Kimi base in our blog from the start. We’ll fix that for the next model." This apology signals a commitment to greater transparency in future product announcements, recognizing the importance of clear communication about model origins and collaborations within the open-source AI ecosystem.

This incident underscores the complex dynamics of AI development, where open-source collaboration intersects with commercial interests, competitive pressures, and geopolitical sensitivities. While Cursor’s use of Kimi 2.5 was ultimately deemed legitimate and supported by Moonshot AI, the initial lack of transparency highlights the ongoing challenges companies face in navigating public perception and competitive landscapes in the rapidly evolving world of artificial intelligence.

TechCrunch Event

San Francisco, CA | October 13-15, 2026

About the Author

Anthony Ha is TechCrunch’s weekend editor. Previously, he worked as a tech reporter at Adweek, a senior editor at VentureBeat, a local government reporter at the Hollister Free Lance, and vice president of content at a VC firm. He lives in New York City. You can contact or verify outreach from Anthony by emailing [email protected].