Cursor AI Hid a Big Secret: Its New Coding Model Is Built on Chinese AI Kimi
Cursor AI launched what it called "frontier-level coding intelligence" this week. But within hours, the tech community uncovered something the company never bothered to mention — its flashy new model, Composer 2, is built on top of Kimi 2.5, an open-source AI model created by a Chinese company. Here is everything you need to know about the controversy, and why it matters far beyond one startup's product launch.
 |
| Credit: Cursor |
What Is Composer 2 and Why Did Cursor Launch It?
Cursor is an AI-powered coding tool that has taken the developer world by storm. The company raised a staggering funding round last fall, pulling in 2.3 billion dollars at a valuation of 29.3 billion dollars. It is also reportedly surpassing 2 billion dollars in annualized revenue — numbers that place it firmly among the most valuable AI startups in the world.
This week, Cursor unveiled Composer 2 with bold claims about its capabilities. The company marketed it as a step forward in AI-assisted coding, promising developers a smarter, faster, and more capable experience. The announcement was enthusiastic and detailed — but it left out one very important piece of information.
There was no mention of Kimi. There was no mention of Moonshot AI. There was certainly no mention of China.
How a Single Developer Exposed the Truth
The story unraveled quickly. A developer posting online under the name Fynn spotted something suspicious inside the model's code. The code appeared to identify Kimi as the underlying model powering Composer 2.
Fynn's reaction was blunt. The post pointed out that Cursor had not even bothered to rename the model identifier in its own codebase. The implication was clear — this was not an original model built from the ground up. It was Kimi 2.5 with additional training layered on top.
The post spread fast in developer circles, forcing Cursor to respond publicly. What followed was a damage-control conversation that raised more questions than it answered.
What Is Kimi 2.5 and Who Made It?
Kimi 2.5 is an open-source AI model released by Moonshot AI, a Chinese artificial intelligence company backed by Alibaba and investment firm HongShan, formerly known as Sequoia China. The model is publicly available, meaning other companies can legally build on top of it under the right licensing terms.
Open-source AI models have become a competitive battleground. Companies release powerful models publicly, and others use them as foundations to build specialized tools. This practice is common and, when done correctly, entirely above board. The problem in Cursor's case was not the use of Kimi — it was the silence around it.
Moonshot AI itself later congratulated Cursor publicly, saying it was proud to see Kimi serve as the foundation for Composer 2. The company described the arrangement as part of an authorized commercial partnership. But none of that transparency came from Cursor at launch.
Cursor's VP Responds — and Defends the Decision
After the story broke, Cursor's vice president of developer education stepped in to address the backlash. He confirmed that Composer 2 did indeed start from an open-source base, acknowledging what the code had already revealed.
However, he pushed back on the idea that Composer 2 is simply a rebranded version of Kimi. According to his explanation, only about one quarter of the computing power used to create the final model came from the Kimi base. The remaining three quarters came from Cursor's own training work. He argued this makes the performance profile of Composer 2 meaningfully different from the original Kimi 2.5 on standard benchmarks.
He also confirmed that Cursor's use of Kimi was fully consistent with the terms of its open-source license. Moonshot AI backed this claim, noting the integration was done through an authorized commercial partnership facilitated by AI infrastructure company Fireworks AI.
The technical defense may hold up. But the communication failure is harder to explain away.
Why Did Cursor Stay Silent About Kimi?
This is the question the tech industry is asking most loudly. Cursor co-founder Aman Sanger eventually admitted the company made a mistake. His statement was straightforward — it was a miss not to mention the Kimi base model from the beginning, and the company would correct that approach for future model launches.
But why stay silent in the first place? The answer likely involves a combination of competitive optics and geopolitical sensitivity.
Building on a Chinese AI model is not illegal, unethical, or even unusual in the open-source world. But optics matter enormously in the current climate. The AI industry has spent the past year framing artificial intelligence development as a strategic competition between the United States and China. When Chinese company DeepSeek released a highly competitive model earlier in 2025, it triggered a wave of concern across Silicon Valley. Several prominent figures described it as a wake-up call for American AI leadership.
Against that backdrop, a well-funded American AI startup quietly building its flagship model on top of a Chinese foundation model is a story that writes itself — and not in a flattering way for Cursor.
Open-Source AI and the US-China Rivalry
Cursor's situation highlights a tension that is only going to grow more complicated. The open-source AI ecosystem does not respect national borders. Models built in China get used by American companies. Models built in America get used by Chinese companies. That is the nature of open collaboration in technology.
But the political framing around AI has become intensely nationalistic. Policymakers, investors, and media figures regularly describe AI development as a zero-sum competition between superpowers. In that environment, even perfectly legal and technically sound decisions can become public relations disasters if companies are not upfront about them.
Cursor's mistake was not using Kimi. Its mistake was treating transparency as optional in a moment when the public — and developers especially — are paying close attention to exactly these kinds of decisions.
What This Means for Developers Using Cursor
For the developers who rely on Cursor daily, the practical question is straightforward. Does knowing that Composer 2 is built on Kimi 2.5 change how they feel about using it?
The honest answer depends on what they were expecting. If a developer assumed Cursor was building all of its AI capabilities from scratch, that assumption was never explicitly promised — but it was implicitly encouraged by marketing language that spoke of frontier intelligence without crediting any external foundation.
On the technical side, the model still needs to perform. Cursor's claim that its additional training substantially changed the model's behavior is testable, and the developer community will be running those tests rigorously in the coming days. If Composer 2 genuinely outperforms the base Kimi model by a meaningful margin, much of the criticism will soften.
If it does not, the controversy will deepen.
A Lesson in Transparency for the AI Industry
Beyond Cursor specifically, this episode is a case study in how not to handle open-source attribution. The AI industry builds constantly on the work of others. That is a feature, not a bug. But when companies benefit from that ecosystem while obscuring their dependencies — especially when those dependencies carry geopolitical weight — they invite exactly the kind of backlash Cursor is now experiencing.
The path forward is not complicated. Name your foundations. Credit your sources. Let the technical work speak for itself. Developers are sophisticated enough to understand that using an open-source base model is standard practice. They are also sophisticated enough to notice when a company is being evasive about it.
Cursor has now committed to doing better next time. Given its size, its funding, and its position in the developer ecosystem, holding it to that promise seems entirely reasonable.
