Mindgard

Chinese AI Models Allegedly Bypassed Safety Limits to Explain Bioweapon Methods, Researchers Say

A U.S.-based AI safety group says it found that two Chinese-language models from Kimi could be pushed past their safeguards and induced to provide instructions related to biological weapons. The finding intensifies global concern that advanced chatbots may be vulnerable to jailbreaks that expose dangerous dual-use knowledge, even when developers claim strong safety controls.