Mindgard researchers uncovered critical safety gaps in Kimi K2.6 and K3 Swarm AI models, revealing how to crea...