UK AISI and CAISI publish first joint cyber assessment of Kimi K3

The British (AISI) and American (CAISI) AI security institutes publish their first joint preliminary assessment of Kimi K3's cyber capabilities. Strong signal: open-weight frontier models fall within the scope of government evaluation.

The fact UK AISI and US CAISI publish their first joint preliminary assessment of Kimi K3's cyber capabilities (Moonshot AI, 2.8 T parameters, GA on July 16, 2026). The kimi-k3-launch thread gains an anticipated government chapter.

Our analysis Two signals. One - the joint assessment shows that both institutes align their methodologies to cover a Chinese open-weight model - unusual in a context where the US axis leans towards controlled access (thread frontier-access-control). Two - the assessment is said to be « preliminary »: it's a methodological framework in the process of being established, more than an operational verdict. What matters is that the format is being standardized before the next frontier open models (Qwen 3.8 Max, future steps by Moonshot) arrive.

To watch Publication of the detailed report, timeline for the Qwen 3.8 Max assessment, propagation of the methodology on the EU side (Brussels / AI Office), and articulation with Hassabis' proposal for an independent regulator (« FINRA of AI », publi #1116).