China, 30 September 2026: Chinese AI developer Moonshot toh conduct kuri ase internal review researches khan duita popular Kimi models ke persuade kuribo para pichideh taikhan ke kubi leh kina biological weapons banabo aro assassinations khan kuribo. Mindgard, kun toh security test kureh AI systems laka, BBC ke koishey taikhan jani paishey July te ki Kimi K2.6 aro K3 Swarm toh safety limits evade kuribo pareh developers pra hala. Olai shey process time te ‘jailbreaking’ laka kute researchers khan complex instructions chulai sabole ini AI khan guardrails ignore kureh nanai.
Moonshot pra BBC ke koishey taikhan third-party input welcomed kuri shey key pillar hoina better aro safer AI banabo leh. Company pra aro koishey BBC ke ki taikhan discussion te ase Mindgard lokot findings opor te Mindgrad laka founder Peter Garraghan pra BBC World Service program Tech Life ke koishey ki findings toh Kimi K2.6 aro K3 Swarm laka concerning ase koina.
Chinese Ai tool researchers ke koishey kina bioweapons banabo
