Chinese AI Models Bypassed Safety Limits on Weapons
Chinese AI tool told researchers how to make bioweapons
Chinese AI developer Moonshot is conducting an internal review after security researchers discovered that two versions of its Kimi chatbot could be manipulated into providing instructions for creating biological weapons and planning assassinations, according to reporting by the BBC.
Mindgard, a firm that tests AI system security, identified the vulnerability in July involving Kimi K2.6 and K3 Swarm models. Researchers used a technique called "jailbreaking"—employing complex instructions to probe whether AI tools ignore safety guardrails—and found that Kimi's safety protections could be circumvented.
The discovery highlights ongoing challenges in AI safety as developers work to prevent their systems from assisting with harmful activities. Moonshot's internal review will examine how the models' safeguards failed and what corrections are needed.
