Russian Hackers Manipulate AI Safety Filters in Cyberattacks on Ukraine

Russian state hackers from the group UAC-0099 embedded a manipulative prompt in a malicious VBS script to trigger AI safety mechanisms, disrupting AI-assisted malware analysis in Ukraine. The technique, named GuardBreaker by ESET, was observed in operations where UAC-0099 conducts initial-access attacks and hands targets to GRU-linked Sandworm hackers. The prompt was hidden as comments within the VBS script. ESET identified the activity as part of ongoing cyber operations targeting Ukraine.