State-backed hackers are deploying malware that uses large language models to dynamically generate malicious scripts, enhancing their ability to evade detection. This development marks a significant advancement towards more autonomous and adaptive malware, with threat actors already experimenting with AI-driven tactics. #PROMPTFLUX #PROMPTSTEAL #FROZENLAKE #BlueDelta #AIthreats
Keypoints
- For the first time, malware using large language models has been observed during execution to alter behavior dynamically.
- Researchers identified malware like PROMPTFLUX and PROMPTSTEAL, which utilize AI to evade detection and generate commands.
- Threat actors, including Russia-linked APT28, are testing AI-enabled malware in operational environments against Ukrainian targets.
- Such AI-driven methods represent a move beyond simple hash-hitting or static scripting, toward more autonomous attacks.
- The underground market for AI tools tailored for cybercrime is expanding, making advanced attacks more accessible to amateurs and professionals alike.
Read More: https://therecord.media/new-malware-uses-ai-to-adapt