AssetID: 56069356
Headline: UNCAPTIONED: Chinese AI Tool Gave Researchers Instructions On Bioweapons
Caption: Chinese AI Tool Gave Researchers Instructions On Bioweapons. Chinese AI developer Moonshot is reviewing its safety systems after researchers said they were able to bypass safeguards on two of its Kimi artificial intelligence models. AI security company Mindgard said tests carried out in July found that Kimi K2.6 and K3 Swarm could be manipulated through so-called “jailbreaking”. Researchers said the models then provided information related to making biological weapons and carrying out assassinations, despite safeguards designed to prevent discussion of harmful activities. Mindgard founder Peter Garraghan said that once the protections were bypassed, the models could provide responses across a range of dangerous topics. Moonshot said it welcomed third-party research as an important part of developing safer AI systems and confirmed it was discussing the findings with Mindgard. The incident has added to concerns about whether safety protections can keep pace with increasingly capable AI models. Instructions: THIS VIDEO MUST NOT BE EDITED FOR LENGTH TO COMBINE WITH OTHER CONTENT
Keywords: Science & Technology,Chinese,AI,Artificial Intelligence,tool,researchers,instructions,bioweapons,build,bypass,safeguards,Kimi,intelligence models,Mindgard security,tests,K2.6,K3 Swarm,manipulated,jailbreaking
PersonInImage: