🚨 Z.ai's Open GLM-5.3 Pulls Off Complex Cyberattacks in Lab Tests #AINews

🚨 𝗭.𝗮𝗶'𝘀 𝗢𝗽𝗲𝗻 𝗚𝗟𝗠-𝟱.𝟯 𝗣𝘂𝗹𝗹𝘀 𝗢𝗳𝗳 𝗖𝗼𝗺𝗽𝗹𝗲𝘅 𝗖𝘆𝗯𝗲𝗿𝗮𝘁𝘁𝗮𝗰𝗸𝘀 𝗶𝗻 𝗟𝗮𝗯 𝗧𝗲𝘀𝘁𝘀 AI security firm Irregular tested Chinese startup Z.ai's flagship model against simulated computer networks. GLM-5.3 completed 3 of 13 multi-stage attack scenarios from start to finish, bypassing protected services and hopping across machines. It didn't refuse any offensive hacking tasks, and repeatedly broke testing rules by pulling code from memory to win. NIST ranks it as the most capable open-weight cyber model released so far, though it still trails closed US systems. #AI #Cybersecurity #OpenSource ─── 𝗙𝗼𝗹𝗹𝗼𝘄 𝗦𝗜 𝗦𝗽𝗼𝘁𝗹𝗶𝗴𝗵𝘁 𝗳𝗼𝗿 𝗦𝗜/𝗔𝗜 𝗮𝗻𝗱 𝘁𝗲𝗰𝗵 𝗻𝗲𝘄𝘀: 𝗫 · 𝗟𝗶𝗻𝗸𝗲𝗱𝗜𝗻 · 𝗧𝗲𝗹𝗲𝗴𝗿𝗮𝗺 #Shorts