Chinese AI tool told researchers how to make bioweapons

Mindgard found that Moonshot's Kimi K2.6 and K3 Swarm models could be jailbroken into discussing biological weapons and assassinations. Moonshot said it is conducting an internal review and is in talks with the researchers, who first alerted it on 27 July and published their findings on 12 September.

• Mindgard discovered the jailbreak in Kimi K2.6 and K3 Swarm in July
• Once jailbroken, the model discusses any topic, including illicit ones
• Moonshot was alerted on 27 July but only responded after BBC's request
• Kimi is an open-weight model, raising misuse risks