Journalism begins where hype ends

,,

We can only see a short distance ahead, but we can see plenty there that needs to be done.” "

—Alan Turing

China’s Kimi K3 Lags Top U.S. Frontier AI Models in Cybersecurity Tests: US CAISI, UK AISI

Illustration showing China's Kimi K3 alongside Anthropic's Claude Fable and OpenAI's Sol, the frontier AI models referenced in the UK-U.S. cybersecurity evaluation.

The preliminary US-UK joint assessment found Kimi K3 performed below the most recent frontier cyber-capable models on several offensive cybersecurity tasks and failed to achieve arbitrary code execution in ExploitBench tests. The evaluation comes amid growing U.S. scrutiny of China’s latest open-weight AI model.