Security researchers discovered in July that Chinese AI developer Moonshot's Kimi K2.6 and K3 Swarm models could be jailbroken to provide instructions for making biological weapons and assassinations. Moonshot is conducting an internal review and discussing the findings with Mindgard.