China’s Moonshot AI challenges US tech giants with new Kimi K3 model

Chinese AI company Moonshot seems ready to take on Western tech leaders in the race for artificial intelligence dominance, following the release of its latest model.

Chinese AI company Moonshot seems ready to take on Western tech leaders in the race for artificial intelligence dominance, following the release of its latest model.

The company recently launched Kimi K3, an open-source AI system reportedly matching the performance of leading US models from OpenAI and Anthropic. Beyond its capabilities, the model’s biggest draw may be its affordability. Independent analysis firm Artificial Analysis found that running a benchmark task through Kimi K3 costs around $0.95, compared to $1.04 for comparable US-based systems.

Moonshot’s decision to make the model fully open-source could pose a serious challenge to American dominance in the space, since it allows users to freely download and modify the underlying code without restrictions, a stark contrast to the more tightly controlled approach taken by most US firms.

Adding to the model’s appeal, Kimi K3 has reportedly been able to resolve serious security vulnerabilities that other AI systems, including OpenAI’s Codex and Anthropic’s Fable, declined to fix due to built-in safety restrictions. This distinction caught the attention of David Sacks, Chair of the President’s Council of Advisors on Science and Technology, who took to X to criticize the trend. He pointed out that Kimi K3 had resolved 15 critical security bugs that American models refused to touch over so-called cyber guardrails, arguing there was little justification for restricting US models on tasks Chinese systems were handling without hesitation, and warning that such caution risked making American AI less competitive globally.

The rise of a cheaper, high-performing alternative also threatens the business models of major US AI labs, many of which depend on steep subscription pricing to offset heavy infrastructure costs.

Alan Woodward, a cybersecurity expert at the University of Surrey, noted that China appears to be taking a more open approach by releasing its models publicly, while the US continues trying to limit their spread. He suggested this strategic difference could eventually allow China to gain the upper hand, potentially leading to tighter controls from Chinese developers once that dominance is established.

However, not everyone views the open-source approach positively. Some experts have raised concerns that the same openness making Kimi K3 attractive could also make it harder to contain, warning that its potential for misuse in cyberattacks may be far more difficult for authorities to regulate compared to the more restricted, proprietary systems used by US companies.