U.S. markets were rattled by Chinese startup Moonshot AI’s July 16 release of Kimi K3, a 2.8 trillion-parameter open-weight model that rivals premium U.S. frontier models like OpenAI‘s GPT-5.6 and Anthropic‘s Claude Fable 5. The release has made headlines across the world and conjured a range of reactions, with some questioning if the US is still leading in the AI race.
The difference between American closed-system labs and Chinese open-weight systems has been thrown into stark relief with the release of Kimi K3. While American AI labs still place higher in overall benchmarks, Kimi K3 has closed the gap so significantly that there’s little daylight between them. It’s also a fraction of the cost.
With the release of Kimi K3, American AI companies are in a panic. As evidenced by a whipsawing stock market, folks are wondering if Kimi K3 is a repeat of DeepSeek’s release of its R1 model in January 2025, or a sign of a more lasting shift in the AI race.
But what about Europe? Despite the European Union’s efforts to develop European-native AI and the required compute capacity, Europeans remain largely dependent on American- and Chinese-made AI.
Historically, American AI has offered the most advanced models, while Chinese AI has been more affordable and adaptable. European-native AI lives somewhere in the middle.
Now, with the release of Kimi K3, a Chinese AI model that’s nearly as capable as its American competitors while being significantly cheaper, European-native AI firms are at a crossroads. Can they compete? Or, are these market conditions temporary?
A Question of Access
In mid-June, the U.S. government temporarily imposed export controls on Anthropic’s Fable 5 and Mythos 5, shutting out Europe and the rest of the world from the most advanced AI models in the world. It’s been rumored that China is considering shuttering access to its most powerful AI models for similar national security reasons.
If these kinds of export controls reemerge, whether temporary or permanent, Europe will need to reckon with its dependency on foreign AI models. The EU’s AI Act is the strictest and most comprehensive AI law in the world. But the U.S. and China did not develop their models by following European rules. European-native AI is competing on an uneven playing field and has been the whole time.
Knowledge Distillation…It’s like Content Scraping…Only Different
Lately, Anthropic and OpenAI have been crying foul regarding a certain practice of China’s AI labs, claiming that the Chinese have stolen the work of American companies. No, OpenAI’s source code hasn’t been hacked. Instead, Chinese firms are methodically scraping the American AI models’ output and reverse engineering the code.
The technique is called Knowledge Distillation, in which a smaller AI model trains on the output of a larger one and mimics what it’s learned. Lo and behold, Anthropic has named Moonshot as one of the labs it detected distilling Anthropic’s models on an industrial scale.
Knowledge Distillation, as a practice, is utilized in Europe too, but is heavily regulated by the EU’s AI Act, and not utilized at the scale that Anthropic alleges Moonshot and other Chinese firms have used it. This leaves European-native AI firms at a competitive disadvantage. Despite the possibility that Moonshot lifted a portion of its code from Anthropic, it’s now a major player in the market and even aims to go public within six months.
The U.S. and China Vie for AI Supremacy While Europe Largely Watches
If you ask Jensen Huang about Moonshot, CEO of NVIDIA, there’s nothing to worry about. Kimi K3 is “excellent” and should be embraced.
Huang’s contention is essentially that there is enough room for everybody at the party. Demand is extensive and diversified enough for both American closed-model AI and lower-cost open-weight Chinese AI to thrive.
Additionally, Huang argues that security concerns in the US about Chinese AI are overblown. In his telling, given the malleable nature of open source AI, American customers could modify the code to protect themselves from China gaining a backdoor.
Whether Huang’s opinions are correct or not, it’s worth noting he is the CEO of the most valuable company in the world, a company which designs and sells the most advanced microchips powering AI. However, China has said ‘no thank you’ to Nvidia’s microchips and focused on procuring domestically produced chips.
Undoubtedly, Huang would be happy if China changed its mind. And therein lies the problem. How much can Huang’s praise of Kimi K3 be taken at face value and how much can it be explained by an ulterior motive to gain market share in China?
Beyond Huang: The State Actors Calling the Real Shots
The same question of ulterior motives could be asked of the primary state actors in the AI race. China and the U.S. are vying for AI hegemony, but going about it very differently. American labs talk of AGI and developing the one AI to rule them all. They argue that government regulation would hamper their ability to beat China. They’ve spent over a trillion dollars in their pursuit.
China’s AI companies, on the other hand, are open source, but heavily regulated by the Communist government. Despite rumors of Beijing throttling access to AI, China has pushed for mass adoption of its version of AI.
Xi Jinping said: “We should seize this rare, historic opportunity to encourage open source, openness, collaboration, and sharing. As a responsible major country, China is always committed to providing international public goods relating to AI.”
On the other hand, OpenAI’s Head of Strategic Futures, Dean Ball, described a world dominated by open weight AI as a “dystopian hellscape” defined by “AI communism.” His opinion isn’t far off from certain members of the Trump Administration.
Increasingly, the AI race is becoming a geopolitical chess game. While Kimi K3 might find a receptive European customer and out-compete European-native AI, Kimi K3’s cutting-edge sophistication could also eventually prompt China to limit foreign access to it.
Where Europe Fits Into a Fight It Didn’t Start
Europe built the world’s strictest AI rulebook while the two countries actually racing for AI supremacy built neither the rules nor the restraint to slow down for it. That gap was survivable when American models simply outclassed the alternatives. It’s less survivable now that a Chinese lab can ship something nearly as good for a fraction of the price, under a government that could cut off access the moment it decides Kimi K3 has become a strategic asset rather than a gift to the world.
Europe doesn’t get to sit this fight out much longer. If Washington and Beijing keep treating access to their best models as leverage, the continent’s dependency stops being an inconvenience and starts being a vulnerability.
If the U.S. institutes another export control on Anthropic and OpenAI, Europe would be in a precarious position. It might be just what it needs to revitalize its push for AI sovereignty.
Author: Tim Tolka, Senior Reporter
See Also:
Every Frontier AI Model Tried to Cheat Its Tests, New AISI Study Finds
When Scientists Built a Fake Society Run Entirely by AI, Grok Turned It Into a Crime Scene
Is Claude Safe for Production? AI Agent Deletes Database in 9 Seconds
